llama.cpp/ggml-vulkan.cpp

7065 lines
384 KiB
C++
Raw Normal View History

ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
#include "ggml-vulkan.h"
#include <vulkan/vulkan_core.h>
#ifdef GGML_VULKAN_RUN_TESTS
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
#include <chrono>
#endif
#include <vulkan/vulkan.hpp>
#include <algorithm>
#include <cmath>
#include <iomanip>
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
#include <iostream>
#include <tuple>
#include <vector>
#include <sstream>
#include <utility>
#include <memory>
#include <limits>
#include <map>
#include <memory>
#include <mutex>
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
#include "ggml.h"
#include "ggml-backend-impl.h"
#include "ggml-vulkan-shaders.hpp"
#define VK_API_VERSION VK_API_VERSION_1_2
#define CEIL_DIV(M, N) (((M) + (N)-1) / (N))
#define VK_VENDOR_ID_AMD 0x1002
#define VK_VENDOR_ID_APPLE 0x106b
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
#define VK_VENDOR_ID_INTEL 0x8086
#define VK_VENDOR_ID_NVIDIA 0x10de
#define VK_DEVICE_DESCRIPTOR_POOL_MODE_UNKNOWN 0
#define VK_DEVICE_DESCRIPTOR_POOL_MODE_MULTI 1
#define VK_DEVICE_DESCRIPTOR_POOL_MODE_SINGLE 2
#define VK_NUM_TYPES 16
#define GGML_VK_MAX_NODES 8192
#define MAX_VK_BUFFERS 256
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
#ifndef K_QUANTS_PER_ITERATION
#define K_QUANTS_PER_ITERATION 1
#else
static_assert(K_QUANTS_PER_ITERATION == 1 || K_QUANTS_PER_ITERATION == 2, "K_QUANTS_PER_ITERATION must be 1 or 2");
#endif
#define VK_CHECK(err, msg) \
do { \
vk::Result err_ = (err); \
if (err_ != vk::Result::eSuccess) { \
fprintf(stderr, "ggml_vulkan: %s error %s at %s:%d\n", \
#err, to_string(err_).c_str(), __FILE__, __LINE__); \
exit(1); \
} \
} while (0)
#ifdef GGML_VULKAN_DEBUG
#define VK_LOG_DEBUG(msg) std::cerr << msg << std::endl
#else
#define VK_LOG_DEBUG(msg) ((void) 0)
#endif // GGML_VULKAN_DEBUG
struct ggml_backend_vk_context;
struct vk_queue {
uint32_t queue_family_index;
vk::Queue queue;
vk::CommandPool pool;
uint32_t cmd_buffer_idx;
std::vector<vk::CommandBuffer> cmd_buffers;
vk::PipelineStageFlags stage_flags;
};
struct vk_pipeline_struct {
std::string name;
vk::ShaderModule shader_module;
vk::DescriptorSetLayout dsl;
std::vector<vk::DescriptorPool> descriptor_pools;
std::vector<vk::DescriptorSet> descriptor_sets;
uint32_t descriptor_set_idx;
vk::PipelineLayout layout;
vk::Pipeline pipeline;
uint32_t push_constant_size;
uint32_t parameter_count;
std::array<uint32_t, 3> wg_denoms;
uint32_t align;
};
typedef std::shared_ptr<vk_pipeline_struct> vk_pipeline;
typedef std::weak_ptr<vk_pipeline_struct> vk_pipeline_ref;
static void ggml_vk_destroy_pipeline(vk::Device& device, vk_pipeline& pipeline);
struct vk_matmul_pipeline_struct {
vk_pipeline l, m, s;
vk_pipeline a_l, a_m, a_s;
};
typedef std::shared_ptr<vk_matmul_pipeline_struct> vk_matmul_pipeline;
struct vk_device_struct;
typedef std::shared_ptr<vk_device_struct> vk_device;
typedef std::weak_ptr<vk_device_struct> vk_device_ref;
struct vk_buffer_struct;
typedef std::shared_ptr<vk_buffer_struct> vk_buffer;
typedef std::weak_ptr<vk_buffer_struct> vk_buffer_ref;
struct ggml_backend_vk_buffer_type_context {
std::string name;
vk_device device;
};
GGML_CALL static const char * ggml_backend_vk_buffer_type_name(ggml_backend_buffer_type_t buft);
GGML_CALL static ggml_backend_buffer_t ggml_backend_vk_buffer_type_alloc_buffer(ggml_backend_buffer_type_t buft, size_t size);
GGML_CALL static size_t ggml_backend_vk_buffer_type_get_alignment(ggml_backend_buffer_type_t buft);
GGML_CALL static size_t ggml_backend_vk_buffer_type_get_max_size(ggml_backend_buffer_type_t buft);
GGML_CALL static size_t ggml_backend_vk_buffer_type_get_alloc_size(ggml_backend_buffer_type_t buft, const ggml_tensor * tensor);
static ggml_backend_buffer_type_i ggml_backend_vk_buffer_type_interface = {
/* .get_name = */ ggml_backend_vk_buffer_type_name,
/* .alloc_buffer = */ ggml_backend_vk_buffer_type_alloc_buffer,
/* .get_alignment = */ ggml_backend_vk_buffer_type_get_alignment,
/* .get_max_size = */ ggml_backend_vk_buffer_type_get_max_size,
/* .get_alloc_size = */ ggml_backend_vk_buffer_type_get_alloc_size,
/* .is_host = */ NULL,
};
#ifdef GGML_VULKAN_MEMORY_DEBUG
class vk_memory_logger;
#endif
static void ggml_vk_destroy_buffer(vk_buffer& buf);
struct vk_device_struct {
std::mutex mutex;
vk::PhysicalDevice physical_device;
vk::PhysicalDeviceProperties properties;
std::string name;
uint64_t max_memory_allocation_size;
bool fp16;
vk::Device device;
uint32_t vendor_id;
vk_queue compute_queue;
vk_queue transfer_queue;
bool single_queue;
uint32_t descriptor_set_mode;
uint32_t subgroup_size;
bool uma;
size_t idx;
vk_matmul_pipeline pipeline_matmul_f32;
vk_matmul_pipeline pipeline_matmul_f32_f16;
vk_matmul_pipeline pipeline_matmul_f16;
vk_matmul_pipeline pipeline_matmul_f16_f32;
vk_pipeline pipeline_matmul_split_k_reduce;
vk_matmul_pipeline pipeline_dequant_mul_mat_mat[VK_NUM_TYPES];
vk_matmul_pipeline pipeline_matmul_id_f32;
vk_matmul_pipeline pipeline_matmul_id_f16;
vk_matmul_pipeline pipeline_matmul_id_f16_f32;
vk_matmul_pipeline pipeline_dequant_mul_mat_mat_id[VK_NUM_TYPES];
vk_pipeline pipeline_dequant[VK_NUM_TYPES];
vk_pipeline pipeline_dequant_mul_mat_vec_f32_f32[VK_NUM_TYPES];
vk_pipeline pipeline_dequant_mul_mat_vec_f16_f32[VK_NUM_TYPES];
vk_pipeline pipeline_dequant_mul_mat_vec_id_f32[VK_NUM_TYPES];
vk_pipeline pipeline_mul_mat_vec_p021_f16_f32;
vk_pipeline pipeline_mul_mat_vec_nc_f16_f32;
vk_pipeline pipeline_get_rows[VK_NUM_TYPES];
vk_pipeline pipeline_get_rows_f32[VK_NUM_TYPES];
vk_pipeline pipeline_mul_f32;
vk_pipeline pipeline_div_f32;
vk_pipeline pipeline_add_f32;
vk_pipeline pipeline_scale_f32;
vk_pipeline pipeline_sqr_f32;
vk_pipeline pipeline_clamp_f32;
vk_pipeline pipeline_cpy_f32_f32, pipeline_cpy_f32_f16, pipeline_cpy_f16_f16;
vk_pipeline pipeline_norm_f32;
vk_pipeline pipeline_rms_norm_f32;
vk_pipeline pipeline_gelu_f32;
vk_pipeline pipeline_silu_f32;
vk_pipeline pipeline_relu_f32;
vk_pipeline pipeline_diag_mask_inf_f32;
vk_pipeline pipeline_soft_max_f32, pipeline_soft_max_f32_f16;
vk_pipeline pipeline_rope_norm_f32, pipeline_rope_norm_f16;
vk_pipeline pipeline_rope_neox_f32, pipeline_rope_neox_f16;
vk_pipeline pipeline_argsort_f32;
vk_pipeline pipeline_sum_rows_f32;
std::vector<vk_pipeline_ref> pipelines;
std::vector<std::tuple<void*, size_t, vk_buffer>> pinned_memory;
vk::Fence fence;
vk_buffer sync_staging;
ggml_backend_buffer_type buffer_type;
#ifdef GGML_VULKAN_MEMORY_DEBUG
std::unique_ptr<vk_memory_logger> memory_logger;
#endif
~vk_device_struct() {
VK_LOG_DEBUG("destroy device " << name);
device.destroyFence(fence);
ggml_vk_destroy_buffer(sync_staging);
device.destroyCommandPool(compute_queue.pool);
if (!single_queue) {
device.destroyCommandPool(transfer_queue.pool);
}
for (auto& pipeline : pipelines) {
if (pipeline.expired()) {
continue;
}
vk_pipeline pl = pipeline.lock();
ggml_vk_destroy_pipeline(device, pl);
}
pipelines.clear();
device.destroy();
}
};
struct vk_buffer_struct {
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
vk::Buffer buffer;
vk::DeviceMemory device_memory;
vk::MemoryPropertyFlags memory_property_flags;
void * ptr;
size_t size = 0;
vk_device device;
~vk_buffer_struct() {
if (size == 0) {
return;
}
VK_LOG_DEBUG("~vk_buffer_struct(" << buffer << ", " << size << ")");
device->device.freeMemory(device_memory);
device->device.destroyBuffer(buffer);
}
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
};
struct vk_subbuffer {
vk_buffer buffer;
uint64_t offset;
uint64_t size;
};
struct vk_semaphore {
vk::Semaphore s;
uint64_t value;
};
struct vk_submission {
vk::CommandBuffer buffer;
std::vector<vk_semaphore> wait_semaphores;
std::vector<vk_semaphore> signal_semaphores;
};
typedef std::vector<vk_submission> vk_sequence;
struct vk_mat_mat_push_constants {
uint32_t M; uint32_t N; uint32_t K;
uint32_t stride_a; uint32_t stride_b; uint32_t stride_d;
uint32_t batch_stride_a; uint32_t batch_stride_b; uint32_t batch_stride_d;
uint32_t k_split;
uint32_t ne02; uint32_t ne12; uint32_t broadcast2; uint32_t broadcast3;
};
struct vk_mat_vec_push_constants {
uint32_t ncols; uint32_t stride_a; uint32_t stride_b; uint32_t stride_d;
uint32_t batch_stride_a; uint32_t batch_stride_b; uint32_t batch_stride_d;
uint32_t ne02; uint32_t ne12; uint32_t broadcast2; uint32_t broadcast3;
};
struct vk_mat_mat_id_push_constants {
uint32_t M; uint32_t N; uint32_t K;
uint32_t stride_a; uint32_t stride_b; uint32_t stride_d;
uint32_t batch_stride_a; uint32_t batch_stride_b; uint32_t batch_stride_d;
uint32_t nei0; uint32_t nei1; uint32_t nbi1; uint32_t ne11;
};
struct vk_mat_vec_id_push_constants {
uint32_t ncols; uint32_t stride_a; uint32_t stride_b; uint32_t stride_d;
uint32_t batch_stride_a; uint32_t batch_stride_b; uint32_t batch_stride_d;
uint32_t nei0; uint32_t ne11;
};
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
struct vk_op_push_constants {
uint32_t KX;
uint32_t KY;
float param1;
float param2;
};
struct vk_op_unary_push_constants {
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
uint32_t ne;
uint32_t ne00; uint32_t ne01; uint32_t ne02; uint32_t ne03; uint32_t nb00; uint32_t nb01; uint32_t nb02; uint32_t nb03;
uint32_t ne10; uint32_t ne11; uint32_t ne12; uint32_t ne13; uint32_t nb10; uint32_t nb11; uint32_t nb12; uint32_t nb13;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
uint32_t d_offset;
float param1; float param2;
};
struct vk_op_binary_push_constants {
uint32_t ne;
uint32_t ne00; uint32_t ne01; uint32_t ne02; uint32_t ne03; uint32_t nb00; uint32_t nb01; uint32_t nb02; uint32_t nb03;
uint32_t ne10; uint32_t ne11; uint32_t ne12; uint32_t ne13; uint32_t nb10; uint32_t nb11; uint32_t nb12; uint32_t nb13;
uint32_t ne20; uint32_t ne21; uint32_t ne22; uint32_t ne23; uint32_t nb20; uint32_t nb21; uint32_t nb22; uint32_t nb23;
uint32_t d_offset;
float param1; float param2;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
};
struct vk_op_diag_mask_push_constants {
uint32_t ncols;
uint32_t rows_per_channel;
int32_t n_past;
};
struct vk_op_rope_push_constants {
uint32_t ncols;
uint32_t n_dims;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
float freq_scale;
uint32_t p_delta_rows;
float freq_base;
float ext_factor;
float attn_factor;
float corr_dims[2];
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
float theta_scale;
uint32_t has_ff;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
};
struct vk_op_soft_max_push_constants {
uint32_t KX;
uint32_t KY;
float scale;
float max_bias;
float m0;
float m1;
uint32_t n_head_log2;
};
struct vk_op_argsort_push_constants {
uint32_t ncols;
uint32_t ncols_pad;
int32_t order;
};
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
// Allow pre-recording command buffers
struct vk_staging_memcpy {
vk_staging_memcpy(void * _dst, const void * _src, size_t _n) : dst(_dst), src(_src), n(_n) {}
void * dst;
const void * src;
size_t n;
};
struct vk_context {
size_t idx;
vk_submission * s;
std::vector<vk_sequence> seqs;
ggml_tensor * exit_tensor;
std::vector<vk_staging_memcpy> in_memcpys;
std::vector<vk_staging_memcpy> out_memcpys;
vk_queue * q;
};
struct ggml_tensor_extra_gpu {
size_t ctx_idx;
vk_buffer_ref buffer_gpu;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
uint64_t offset;
void reset() {
ctx_idx = 0;
buffer_gpu.reset();
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
offset = 0;
}
};
struct ggml_vk_garbage_collector {
std::vector<vk_semaphore> tl_semaphores;
std::vector<vk_semaphore> semaphores;
std::vector<vk::Event> events;
std::vector<vk_buffer> temp_buffers;
std::vector<vk_context> contexts;
};
#if defined(GGML_VULKAN_MEMORY_DEBUG) || defined(GGML_VULKAN_DEBUG)
#define VK_LOG_MEMORY(msg) std::cerr << "ggml_vulkan memory: " << msg << std::endl
static std::string format_size(size_t size) {
const size_t kib = 1024;
const size_t mib = kib * 1024;
const size_t gib = mib * 1024;
std::ostringstream oss;
oss << std::fixed << std::setprecision(2);
if (size >= gib) {
oss << static_cast<double>(size) / gib << " GiB";
} else if (size >= mib) {
oss << static_cast<double>(size) / mib << " MiB";
} else if (size >= kib) {
oss << static_cast<double>(size) / kib << " KiB";
} else {
oss << size << " B";
}
return oss.str();
}
static std::mutex log_mutex;
class vk_memory_logger {
public:
vk_memory_logger(): total_device(0), total_host(0) {}
void log_allocation(vk_buffer_ref buf_ref, size_t size);
void log_deallocation(vk_buffer_ref buf_ref);
private:
std::map<vk::Buffer, size_t> allocations; // Track allocations
size_t total_device;
size_t total_host;
};
#else
#define VK_LOG_MEMORY(msg) ((void) 0)
#endif // GGML_VULKAN_MEMORY_DEBUG
struct ggml_backend_vk_context {
std::string name;
vk_device device;
size_t semaphore_idx, event_idx;
ggml_vk_garbage_collector gc;
size_t prealloc_size_x, prealloc_size_y, prealloc_size_split_k;
vk_buffer prealloc_x, prealloc_y, prealloc_split_k;
vk::Fence fence;
vk_buffer staging;
size_t staging_size;
size_t staging_offset;
vk_buffer buffer_pool[MAX_VK_BUFFERS];
vk_context * compute_ctx;
vk_context * transfer_ctx;
};
#ifdef GGML_VULKAN_MEMORY_DEBUG
void vk_memory_logger::log_allocation(vk_buffer_ref buf_ref, size_t size) {
std::lock_guard<std::mutex> guard(log_mutex);
vk_buffer buf = buf_ref.lock();
const bool device = bool(buf->memory_property_flags & vk::MemoryPropertyFlagBits::eDeviceLocal);
const std::string type = device ? "device" : "host";
allocations[buf->buffer] = size;
total_device += device ? size : 0;
total_host += device ? 0 : size;
VK_LOG_MEMORY(buf->device->name << ": +" << format_size(size) << " " << type << " at " << buf->buffer << ". Total device: " << format_size(total_device) << ", total host: " << format_size(total_host));
}
void vk_memory_logger::log_deallocation(vk_buffer_ref buf_ref) {
if (buf_ref.expired() || buf_ref.lock()->size == 0) {
return;
}
std::lock_guard<std::mutex> guard(log_mutex);
vk_buffer buf = buf_ref.lock();
const bool device = bool(buf->memory_property_flags & vk::MemoryPropertyFlagBits::eDeviceLocal);
std::string type = device ? "device" : "host";
auto it = allocations.find(buf->buffer);
total_device -= device ? it->second : 0;
total_host -= device ? 0 : it->second;
if (it != allocations.end()) {
VK_LOG_MEMORY(buf->device->name << ": -" << format_size(it->second) << " " << type << " at " << buf->buffer << ". Total device: " << format_size(total_device) << ", total host: " << format_size(total_host));
allocations.erase(it);
} else {
VK_LOG_MEMORY("ERROR " << buf->device->name << ": Attempted to deallocate unknown " << type << " memory at " << buf->buffer);
}
}
#endif // GGML_VULKAN_MEMORY_DEBUG
struct vk_instance_t {
vk::Instance instance;
std::vector<size_t> device_indices;
vk_device devices[GGML_VK_MAX_DEVICES];
};
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
static bool vk_instance_initialized = false;
static vk_instance_t vk_instance;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
#ifdef GGML_VULKAN_CHECK_RESULTS
static size_t vk_skip_checks;
static size_t vk_output_tensor;
static void ggml_vk_print_tensor(ggml_backend * ctx, const ggml_tensor * tensor, const char * name);
static void ggml_vk_check_results_0(ggml_backend_vk_context * ctx, ggml_tensor * tensor);
static void ggml_vk_check_results_1(ggml_backend_vk_context * ctx, ggml_tensor * tensor);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
#endif
typedef void (*ggml_vk_func_t)(ggml_backend_vk_context * ctx, vk_context * subctx, const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst);
GGML_CALL static void ggml_backend_vk_free(ggml_backend_t backend);
static void ggml_vk_create_pipeline(vk_device& device, vk_pipeline& pipeline, const std::string& name, size_t spv_size, const void* spv_data, const std::string& entrypoint, uint32_t parameter_count, uint32_t push_constant_size, std::array<uint32_t, 3> wg_denoms, std::vector<uint32_t>&& specialization_constants, uint32_t align) {
VK_LOG_DEBUG("ggml_vk_create_pipeline(" << device->name << ", " << name << ", " << entrypoint << ", " << parameter_count << ", " << push_constant_size << ", (" << wg_denoms[0] << "," << wg_denoms[1] << "," << wg_denoms[2] << "), specialization_constants, " << align << ")");
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
GGML_ASSERT(parameter_count > 0);
GGML_ASSERT(wg_denoms[0] > 0 && wg_denoms[1] > 0 && wg_denoms[2] > 0); // NOLINT
std::lock_guard<std::mutex> guard(device->mutex);
pipeline = std::make_shared<vk_pipeline_struct>();
pipeline->name = name;
pipeline->parameter_count = parameter_count;
pipeline->push_constant_size = push_constant_size;
pipeline->wg_denoms = wg_denoms;
pipeline->align = align;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
vk::ShaderModuleCreateInfo shader_module_create_info({}, spv_size, reinterpret_cast<const uint32_t *>(spv_data));
pipeline->shader_module = device->device.createShaderModule(shader_module_create_info);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
std::vector<vk::DescriptorSetLayoutBinding> dsl_binding;
std::vector<vk::DescriptorBindingFlags> dsl_binding_flags;
for (uint32_t i = 0; i < parameter_count; i++) {
dsl_binding.push_back({i, vk::DescriptorType::eStorageBuffer, 1, vk::ShaderStageFlagBits::eCompute});
dsl_binding_flags.push_back({});
}
vk::DescriptorSetLayoutBindingFlagsCreateInfo dslbfci = { dsl_binding_flags };
vk::PushConstantRange pcr(
vk::ShaderStageFlagBits::eCompute,
0,
pipeline->push_constant_size
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
);
vk::DescriptorSetLayoutCreateInfo descriptor_set_layout_create_info(
{},
dsl_binding);
descriptor_set_layout_create_info.setPNext(&dslbfci);
pipeline->dsl = device->device.createDescriptorSetLayout(descriptor_set_layout_create_info);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
// Check if device supports multiple descriptors per pool
if (device->descriptor_set_mode == VK_DEVICE_DESCRIPTOR_POOL_MODE_UNKNOWN) {
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
const uint32_t alloc_count = 2;
// Try allocating multiple sets from one pool
// This fails on AMD for some reason, so add a fall back to allocating one pool per set
vk::DescriptorPoolSize descriptor_pool_size(vk::DescriptorType::eStorageBuffer, pipeline->parameter_count);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
vk::DescriptorPoolCreateInfo descriptor_pool_create_info({}, alloc_count, descriptor_pool_size);
vk::DescriptorPool pool = device->device.createDescriptorPool(descriptor_pool_create_info);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
std::vector<vk::DescriptorSetLayout> layouts(alloc_count);
for (uint32_t i = 0; i < alloc_count; i++) {
layouts[i] = pipeline->dsl;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
try {
vk::DescriptorSetAllocateInfo descriptor_set_alloc_info(pool, alloc_count, layouts.data());
std::vector<vk::DescriptorSet> sets = device->device.allocateDescriptorSets(descriptor_set_alloc_info);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
} catch(vk::OutOfPoolMemoryError const&) {
device->descriptor_set_mode = VK_DEVICE_DESCRIPTOR_POOL_MODE_SINGLE;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
device->device.destroyDescriptorPool(pool);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
if (device->descriptor_set_mode == VK_DEVICE_DESCRIPTOR_POOL_MODE_MULTI) {
vk::DescriptorPoolSize descriptor_pool_size(vk::DescriptorType::eStorageBuffer, pipeline->parameter_count);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
vk::DescriptorPoolCreateInfo descriptor_pool_create_info({}, 128, descriptor_pool_size);
pipeline->descriptor_pools.push_back(device->device.createDescriptorPool(descriptor_pool_create_info));
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
pipeline->descriptor_set_idx = 0;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
vk::PipelineLayoutCreateInfo pipeline_layout_create_info(vk::PipelineLayoutCreateFlags(), pipeline->dsl, pcr);
pipeline->layout = device->device.createPipelineLayout(pipeline_layout_create_info);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
std::vector<vk::SpecializationMapEntry> specialization_entries(specialization_constants.size());
for (size_t i = 0; i < specialization_constants.size(); i++) {
specialization_entries[i].constantID = i;
specialization_entries[i].offset = i * sizeof(uint32_t);
specialization_entries[i].size = sizeof(uint32_t);
}
vk::SpecializationInfo specialization_info(
specialization_entries.size(),
specialization_entries.data(),
specialization_constants.size() * sizeof(uint32_t),
specialization_constants.data()
);
vk::PipelineShaderStageCreateInfo pipeline_shader_create_info(
vk::PipelineShaderStageCreateFlags(),
vk::ShaderStageFlagBits::eCompute,
pipeline->shader_module,
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
entrypoint.c_str(),
&specialization_info);
vk::ComputePipelineCreateInfo compute_pipeline_create_info(
vk::PipelineCreateFlags(),
pipeline_shader_create_info,
pipeline->layout);
pipeline->pipeline = device->device.createComputePipeline(VK_NULL_HANDLE, compute_pipeline_create_info).value;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
device->pipelines.push_back(pipeline);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
static void ggml_vk_destroy_pipeline(vk::Device& device, vk_pipeline& pipeline) {
VK_LOG_DEBUG("ggml_pipeline_destroy_pipeline(" << pipeline->name << ")");
for (auto& pool : pipeline->descriptor_pools) {
device.destroyDescriptorPool(pool);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
pipeline->descriptor_pools.clear();
pipeline->descriptor_sets.clear();
pipeline->descriptor_set_idx = 0;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
device.destroyDescriptorSetLayout(pipeline->dsl);
device.destroyPipelineLayout(pipeline->layout);
device.destroyShaderModule(pipeline->shader_module);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
device.destroyPipeline(pipeline->pipeline);
}
static void ggml_pipeline_allocate_descriptor_sets(vk_device& device, vk_pipeline& pipeline, uint32_t n) {
VK_LOG_DEBUG("ggml_pipeline_allocate_descriptor_sets(" << pipeline->name << ", " << n << ")");
if (pipeline->descriptor_sets.size() >= pipeline->descriptor_set_idx + n) {
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
// Enough descriptors are available
return;
}
std::lock_guard<std::mutex> guard(device->mutex);
if (device->descriptor_set_mode == VK_DEVICE_DESCRIPTOR_POOL_MODE_MULTI) {
const uint32_t alloc_count = pipeline->descriptor_set_idx + n - pipeline->descriptor_sets.size();
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
std::vector<vk::DescriptorSetLayout> layouts(alloc_count);
for (uint32_t i = 0; i < alloc_count; i++) {
layouts[i] = pipeline->dsl;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
vk::DescriptorSetAllocateInfo descriptor_set_alloc_info(pipeline->descriptor_pools[0], alloc_count, layouts.data());
std::vector<vk::DescriptorSet> sets = device->device.allocateDescriptorSets(descriptor_set_alloc_info);
pipeline->descriptor_sets.insert(pipeline->descriptor_sets.end(), sets.begin(), sets.end());
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
} else {
for (uint32_t i = pipeline->descriptor_sets.size(); i < pipeline->descriptor_set_idx + n; i++) {
vk::DescriptorPoolSize descriptor_pool_size(vk::DescriptorType::eStorageBuffer, pipeline->parameter_count);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
vk::DescriptorPoolCreateInfo descriptor_pool_create_info({}, 1, descriptor_pool_size);
pipeline->descriptor_pools.push_back(device->device.createDescriptorPool(descriptor_pool_create_info));
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
vk::DescriptorSetAllocateInfo descriptor_set_alloc_info(pipeline->descriptor_pools[i], 1, &pipeline->dsl);
std::vector<vk::DescriptorSet> sets = device->device.allocateDescriptorSets(descriptor_set_alloc_info);
pipeline->descriptor_sets.push_back(sets[0]);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
}
}
static void ggml_pipeline_cleanup(vk_pipeline& pipeline) {
VK_LOG_DEBUG("ggml_pipeline_cleanup(" << pipeline->name << ")");
pipeline->descriptor_set_idx = 0;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
static vk::CommandBuffer ggml_vk_create_cmd_buffer(vk_device& device, vk_queue& q) {
VK_LOG_DEBUG("ggml_vk_create_cmd_buffer()");
std::lock_guard<std::mutex> guard(device->mutex);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
if (q.cmd_buffers.size() > q.cmd_buffer_idx) {
// Reuse command buffer
return q.cmd_buffers[q.cmd_buffer_idx++];
}
vk::CommandBufferAllocateInfo command_buffer_alloc_info(
q.pool,
vk::CommandBufferLevel::ePrimary,
1);
const std::vector<vk::CommandBuffer> cmd_buffers = device->device.allocateCommandBuffers(command_buffer_alloc_info);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
auto buf = cmd_buffers.front();
q.cmd_buffers.push_back(buf);
q.cmd_buffer_idx++;
return buf;
}
static vk_submission ggml_vk_create_submission(vk_device& device, vk_queue& q, std::vector<vk_semaphore> wait_semaphores, std::vector<vk_semaphore> signal_semaphores) {
VK_LOG_DEBUG("ggml_vk_create_submission()");
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
vk_submission s;
s.buffer = ggml_vk_create_cmd_buffer(device, q);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
s.wait_semaphores = std::move(wait_semaphores);
s.signal_semaphores = std::move(signal_semaphores);
return s;
}
static void ggml_vk_submit(vk_context * ctx, vk::Fence fence) {
VK_LOG_DEBUG("ggml_vk_submit(" << ctx->seqs.size() << ", " << fence << ")");
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
if (ctx->seqs.empty()) {
return;
}
std::vector<std::vector<uint64_t>> tl_wait_vals;
std::vector<std::vector<uint64_t>> tl_signal_vals;
std::vector<std::vector<vk::Semaphore>> tl_wait_semaphores;
std::vector<std::vector<vk::Semaphore>> tl_signal_semaphores;
std::vector<vk::TimelineSemaphoreSubmitInfo> tl_submit_infos;
std::vector<vk::SubmitInfo> submit_infos;
int idx = -1;
std::vector<std::vector<vk::PipelineStageFlags>> stage_flags;
size_t reserve = 0;
for (const auto& sequence : ctx->seqs) {
reserve += sequence.size();
}
// Pre-reserve vectors to prevent reallocation, which invalidates pointers
tl_wait_semaphores.reserve(reserve);
tl_wait_vals.reserve(reserve);
tl_signal_semaphores.reserve(reserve);
tl_signal_vals.reserve(reserve);
tl_submit_infos.reserve(reserve);
submit_infos.reserve(reserve);
stage_flags.reserve(reserve);
for (const auto& sequence : ctx->seqs) {
for (const auto& submission : sequence) {
stage_flags.push_back({});
idx++;
tl_wait_vals.push_back({});
tl_wait_semaphores.push_back({});
tl_signal_vals.push_back({});
tl_signal_semaphores.push_back({});
for (size_t i = 0; i < submission.wait_semaphores.size(); i++) {
stage_flags[idx].push_back(ctx->q->stage_flags);
tl_wait_vals[idx].push_back(submission.wait_semaphores[i].value);
tl_wait_semaphores[idx].push_back(submission.wait_semaphores[i].s);
}
for (size_t i = 0; i < submission.signal_semaphores.size(); i++) {
tl_signal_vals[idx].push_back(submission.signal_semaphores[i].value);
tl_signal_semaphores[idx].push_back(submission.signal_semaphores[i].s);
}
tl_submit_infos.push_back({
(uint32_t) submission.wait_semaphores.size(),
tl_wait_vals[idx].data(),
(uint32_t) submission.signal_semaphores.size(),
tl_signal_vals[idx].data(),
});
tl_submit_infos[idx].sType = vk::StructureType::eTimelineSemaphoreSubmitInfo;
tl_submit_infos[idx].pNext = nullptr;
vk::SubmitInfo si{
(uint32_t) submission.wait_semaphores.size(),
tl_wait_semaphores[idx].data(),
stage_flags[idx].data(),
1,
&submission.buffer,
(uint32_t) submission.signal_semaphores.size(),
tl_signal_semaphores[idx].data(),
};
si.setPNext(&tl_submit_infos[idx]);
submit_infos.push_back(si);
}
}
ctx->q->queue.submit(submit_infos, fence);
ctx->seqs.clear();
}
static uint32_t ggml_vk_find_queue_family_index(std::vector<vk::QueueFamilyProperties>& queue_family_props, const vk::QueueFlags& required, const vk::QueueFlags& avoid, int32_t compute_index, uint32_t min_num_queues) {
VK_LOG_DEBUG("ggml_vk_find_queue_family_index()");
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
const uint32_t qfsize = queue_family_props.size();
// Try with avoid preferences first
for (uint32_t i = 0; i < qfsize; i++) {
if (queue_family_props[i].queueCount >= min_num_queues && (compute_index < 0 || i != (uint32_t) compute_index) && queue_family_props[i].queueFlags & required && !(queue_family_props[i].queueFlags & avoid)) {
return i;
}
}
// Fall back to only required
for (size_t i = 0; i < qfsize; i++) {
if (queue_family_props[i].queueCount >= min_num_queues && (compute_index < 0 || i != (uint32_t) compute_index) && queue_family_props[i].queueFlags & required) {
return i;
}
}
// Fall back to reusing compute queue
for (size_t i = 0; i < qfsize; i++) {
if (queue_family_props[i].queueCount >= min_num_queues && queue_family_props[i].queueFlags & required) {
return i;
}
}
// Fall back to ignoring min_num_queries
for (size_t i = 0; i < qfsize; i++) {
if (queue_family_props[i].queueFlags & required) {
return i;
}
}
// All commands that are allowed on a queue that supports transfer operations are also allowed on a queue that supports either graphics or compute operations.
// Thus, if the capabilities of a queue family include VK_QUEUE_GRAPHICS_BIT or VK_QUEUE_COMPUTE_BIT, then reporting the VK_QUEUE_TRANSFER_BIT capability separately for that queue family is optional.
if (compute_index >= 0) {
return compute_index;
}
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
std::cerr << "ggml_vulkan: No suitable queue family index found." << std::endl;
for(auto &q_family : queue_family_props) {
std::cerr << "Queue number: " + std::to_string(q_family.queueCount) << " flags: " + to_string(q_family.queueFlags) << std::endl;
}
abort();
}
static void ggml_vk_create_queue(vk_device& device, vk_queue& q, uint32_t queue_family_index, uint32_t queue_index, vk::PipelineStageFlags&& stage_flags) {
VK_LOG_DEBUG("ggml_vk_create_queue()");
std::lock_guard<std::mutex> guard(device->mutex);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
q.queue_family_index = queue_family_index;
vk::CommandPoolCreateInfo command_pool_create_info_compute(vk::CommandPoolCreateFlags(VK_COMMAND_POOL_CREATE_TRANSIENT_BIT), queue_family_index);
q.pool = device->device.createCommandPool(command_pool_create_info_compute);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
q.cmd_buffer_idx = 0;
q.queue = device->device.getQueue(queue_family_index, queue_index);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
q.stage_flags = stage_flags;
}
static vk_context * ggml_vk_create_context(ggml_backend_vk_context * ctx, vk_queue& q) {
VK_LOG_DEBUG("ggml_vk_create_context()");
ctx->gc.contexts.emplace_back();
vk_context * result = &ctx->gc.contexts[ctx->gc.contexts.size() - 1];
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
memset((void *) result, 0, sizeof(vk_context));
result->idx = ctx->gc.contexts.size() - 1;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
result->q = &q;
return result;
}
static vk_context * ggml_vk_create_temporary_context(vk_queue& q) {
VK_LOG_DEBUG("ggml_vk_create_temporary_context()");
vk_context * result = new vk_context;
memset((void *) result, 0, sizeof(vk_context));
result->idx = 0;
result->q = &q;
return result;
}
static vk_semaphore * ggml_vk_create_binary_semaphore(ggml_backend_vk_context * ctx) {
VK_LOG_DEBUG("ggml_vk_create_timeline_semaphore()");
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
vk::SemaphoreTypeCreateInfo tci{ vk::SemaphoreType::eBinary, 0 };
vk::SemaphoreCreateInfo ci{};
ci.setPNext(&tci);
vk::Semaphore semaphore = ctx->device->device.createSemaphore(ci);
ctx->gc.semaphores.push_back({ semaphore, 0 });
return &ctx->gc.semaphores[ctx->gc.semaphores.size() - 1];
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
static vk_semaphore * ggml_vk_create_timeline_semaphore(ggml_backend_vk_context * ctx) {
VK_LOG_DEBUG("ggml_vk_create_timeline_semaphore()");
if (ctx->semaphore_idx >= ctx->gc.tl_semaphores.size()) {
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
vk::SemaphoreTypeCreateInfo tci{ vk::SemaphoreType::eTimeline, 0 };
vk::SemaphoreCreateInfo ci{};
ci.setPNext(&tci);
vk::Semaphore semaphore = ctx->device->device.createSemaphore(ci);
ctx->gc.tl_semaphores.push_back({ semaphore, 0 });
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
return &ctx->gc.tl_semaphores[ctx->semaphore_idx++];
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
static vk::Event ggml_vk_create_event(ggml_backend_vk_context * ctx) {
if (ctx->event_idx >= ctx->gc.events.size()) {
ctx->gc.events.push_back(ctx->device->device.createEvent({}));
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
return ctx->gc.events[ctx->event_idx++];
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
static void ggml_vk_queue_cleanup(vk_device& device, vk_queue& q) {
VK_LOG_DEBUG("ggml_vk_queue_cleanup()");
std::lock_guard<std::mutex> guard(device->mutex);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
// Requires command buffers to be done
device->device.resetCommandPool(q.pool);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
q.cmd_buffer_idx = 0;
}
static uint32_t find_properties(const vk::PhysicalDeviceMemoryProperties* mem_props, vk::MemoryRequirements* mem_req, vk::MemoryPropertyFlags flags) {
for (uint32_t i = 0; i < mem_props->memoryTypeCount; ++i) {
vk::MemoryType memory_type = mem_props->memoryTypes[i];
if ((mem_req->memoryTypeBits & ((uint64_t)1 << i)) &&
(flags & memory_type.propertyFlags) == flags &&
mem_props->memoryHeaps[memory_type.heapIndex].size >= mem_req->size) {
return static_cast<int32_t>(i);
}
}
return UINT32_MAX;
}
static vk_buffer ggml_vk_create_buffer(vk_device& device, size_t size, vk::MemoryPropertyFlags req_flags, vk::MemoryPropertyFlags fallback_flags = vk::MemoryPropertyFlags(0)) {
VK_LOG_DEBUG("ggml_vk_create_buffer(" << device->name << ", " << size << ", " << to_string(req_flags) << ", " << to_string(fallback_flags) << ")");
std::lock_guard<std::mutex> guard(device->mutex);
vk_buffer buf = std::make_shared<vk_buffer_struct>();
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
if (size == 0) {
buf->size = 0;
return buf;
}
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
buf->size = size;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
vk::BufferCreateInfo buffer_create_info{
vk::BufferCreateFlags(),
size,
vk::BufferUsageFlagBits::eStorageBuffer | vk::BufferUsageFlagBits::eTransferSrc | vk::BufferUsageFlagBits::eTransferDst,
vk::SharingMode::eExclusive,
0,
nullptr,
};
buf->buffer = device->device.createBuffer(buffer_create_info);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
vk::MemoryRequirements mem_req = device->device.getBufferMemoryRequirements(buf->buffer);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
vk::PhysicalDeviceMemoryProperties mem_props = device->physical_device.getMemoryProperties();
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
uint32_t memory_type_index = UINT32_MAX;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
memory_type_index = find_properties(&mem_props, &mem_req, req_flags);
buf->memory_property_flags = req_flags;
if (memory_type_index == UINT32_MAX && fallback_flags) {
memory_type_index = find_properties(&mem_props, &mem_req, fallback_flags);
buf->memory_property_flags = fallback_flags;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
if (memory_type_index == UINT32_MAX) {
device->device.destroyBuffer(buf->buffer);
buf->size = 0;
throw vk::OutOfDeviceMemoryError("No suitable memory type found");
}
try {
buf->device_memory = device->device.allocateMemory({ mem_req.size, memory_type_index });
} catch (const vk::SystemError& e) {
// Out of Host/Device memory, clean up buffer
device->device.destroyBuffer(buf->buffer);
buf->size = 0;
throw e;
}
buf->ptr = nullptr;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
if (buf->memory_property_flags & vk::MemoryPropertyFlagBits::eHostVisible) {
buf->ptr = device->device.mapMemory(buf->device_memory, 0, VK_WHOLE_SIZE);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
device->device.bindBufferMemory(buf->buffer, buf->device_memory, 0);
buf->device = device;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
#ifdef GGML_VULKAN_MEMORY_DEBUG
device->memory_logger->log_allocation(buf, size);
#endif
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
return buf;
}
static vk_buffer ggml_vk_create_buffer_check(vk_device& device, size_t size, vk::MemoryPropertyFlags req_flags, vk::MemoryPropertyFlags fallback_flags = vk::MemoryPropertyFlags(0)) {
try {
return ggml_vk_create_buffer(device, size, req_flags, fallback_flags);
} catch (const vk::SystemError& e) {
std::cerr << "ggml_vulkan: Memory allocation of size " << size << " failed." << std::endl;
std::cerr << "ggml_vulkan: " << e.what() << std::endl;
throw e;
}
}
static vk_buffer ggml_vk_create_buffer_device(vk_device& device, size_t size) {
vk_buffer buf;
try {
if (device->uma) {
// Fall back to host memory type
buf = ggml_vk_create_buffer(device, size, vk::MemoryPropertyFlagBits::eDeviceLocal, vk::MemoryPropertyFlagBits::eHostVisible | vk::MemoryPropertyFlagBits::eHostCoherent);
} else {
buf = ggml_vk_create_buffer(device, size, vk::MemoryPropertyFlagBits::eDeviceLocal);
}
} catch (const vk::SystemError& e) {
std::cerr << "ggml_vulkan: Device memory allocation of size " << size << " failed." << std::endl;
std::cerr << "ggml_vulkan: " << e.what() << std::endl;
throw e;
}
return buf;
}
static void ggml_vk_destroy_buffer(vk_buffer& buf) {
if (buf == nullptr) {
return;
}
#ifdef GGML_VULKAN_MEMORY_DEBUG
if (buf->device != nullptr) {
buf->device->memory_logger->log_deallocation(buf);
}
#endif
buf.reset();
}
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
static vk_subbuffer ggml_vk_subbuffer(vk_buffer& buf) {
return { buf, 0, VK_WHOLE_SIZE };
}
static void ggml_vk_sync_buffers(vk_context * ctx) {
VK_LOG_DEBUG("ggml_vk_sync_buffers()");
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
const std::vector<vk::MemoryBarrier> mem_barriers{ { { vk::AccessFlagBits::eMemoryRead | vk::AccessFlagBits::eMemoryWrite }, { vk::AccessFlagBits::eMemoryRead | vk::AccessFlagBits::eMemoryWrite } } };
ctx->s->buffer.pipelineBarrier(
ctx->q->stage_flags,
ctx->q->stage_flags,
{},
mem_barriers,
{},
{}
);
}
static void ggml_vk_wait_events(vk_context * ctx, std::vector<vk::Event>&& events) {
VK_LOG_DEBUG("ggml_vk_wait_events()");
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
if (events.empty()) {
return;
}
ctx->s->buffer.waitEvents(
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
events,
ctx->q->stage_flags,
ctx->q->stage_flags,
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
{},
{},
{}
);
}
static bool ggml_vk_build_shader(ggml_type type) {
switch(type) {
case GGML_TYPE_F16:
case GGML_TYPE_Q4_0:
case GGML_TYPE_Q4_1:
case GGML_TYPE_Q5_0:
case GGML_TYPE_Q5_1:
case GGML_TYPE_Q8_0:
case GGML_TYPE_Q2_K:
case GGML_TYPE_Q3_K:
case GGML_TYPE_Q4_K:
case GGML_TYPE_Q5_K:
case GGML_TYPE_Q6_K:
return true;
default:
return false;
}
}
static void ggml_vk_load_shaders(vk_device& device) {
VK_LOG_DEBUG("ggml_vk_load_shaders(" << device->name << ")");
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
// mulmat
std::initializer_list<uint32_t> warptile_l = { 128, 128, 128, 16, device->subgroup_size * 2, 64, 2, 4, 4, device->subgroup_size };
std::initializer_list<uint32_t> warptile_m = { 128, 64, 64, 16, device->subgroup_size, 32, 2, 4, 2, device->subgroup_size };
std::initializer_list<uint32_t> warptile_s = { device->subgroup_size, 32, 32, 16, 32, 32, 2, 2, 2, device->subgroup_size };
std::initializer_list<uint32_t> warptile_mmq_l = { 128, 128, 128, 32, device->subgroup_size * 2, 64, 2, 4, 4, device->subgroup_size };
std::initializer_list<uint32_t> warptile_mmq_m = { 128, 64, 64, 32, device->subgroup_size, 32, 2, 4, 2, device->subgroup_size };
std::initializer_list<uint32_t> warptile_mmq_s = { device->subgroup_size, 32, 32, 32, 32, 32, 2, 2, 2, device->subgroup_size };
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
std::array<uint32_t, 3> l_wg_denoms = {128, 128, 1 };
std::array<uint32_t, 3> m_wg_denoms = { 64, 64, 1 };
std::array<uint32_t, 3> s_wg_denoms = { 32, 32, 1 };
uint32_t l_align = 128;
uint32_t m_align = 64;
uint32_t s_align = 32;
device->pipeline_matmul_f32 = std::make_shared<vk_matmul_pipeline_struct>();
device->pipeline_matmul_f32_f16 = std::make_shared<vk_matmul_pipeline_struct>();
device->pipeline_matmul_f16_f32 = std::make_shared<vk_matmul_pipeline_struct>();
device->pipeline_matmul_f16 = std::make_shared<vk_matmul_pipeline_struct>();
device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q4_0] = std::make_shared<vk_matmul_pipeline_struct>();
device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q4_1] = std::make_shared<vk_matmul_pipeline_struct>();
device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q5_0] = std::make_shared<vk_matmul_pipeline_struct>();
device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q5_1] = std::make_shared<vk_matmul_pipeline_struct>();
device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q8_0] = std::make_shared<vk_matmul_pipeline_struct>();
device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q2_K] = std::make_shared<vk_matmul_pipeline_struct>();
device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q3_K] = std::make_shared<vk_matmul_pipeline_struct>();
device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q4_K] = std::make_shared<vk_matmul_pipeline_struct>();
device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q5_K] = std::make_shared<vk_matmul_pipeline_struct>();
device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q6_K] = std::make_shared<vk_matmul_pipeline_struct>();
device->pipeline_matmul_id_f32 = std::make_shared<vk_matmul_pipeline_struct>();
device->pipeline_matmul_id_f16_f32 = std::make_shared<vk_matmul_pipeline_struct>();
device->pipeline_matmul_id_f16 = std::make_shared<vk_matmul_pipeline_struct>();
device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q4_0] = std::make_shared<vk_matmul_pipeline_struct>();
device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q4_1] = std::make_shared<vk_matmul_pipeline_struct>();
device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q5_0] = std::make_shared<vk_matmul_pipeline_struct>();
device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q5_1] = std::make_shared<vk_matmul_pipeline_struct>();
device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q8_0] = std::make_shared<vk_matmul_pipeline_struct>();
device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q2_K] = std::make_shared<vk_matmul_pipeline_struct>();
device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q3_K] = std::make_shared<vk_matmul_pipeline_struct>();
device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q4_K] = std::make_shared<vk_matmul_pipeline_struct>();
device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q5_K] = std::make_shared<vk_matmul_pipeline_struct>();
device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q6_K] = std::make_shared<vk_matmul_pipeline_struct>();
if (device->fp16) {
ggml_vk_create_pipeline(device, device->pipeline_matmul_f32->l, "matmul_f32_l", matmul_f32_f32_len, matmul_f32_f32_data, "main", 3, sizeof(vk_mat_mat_push_constants), l_wg_denoms, warptile_l, 1);
ggml_vk_create_pipeline(device, device->pipeline_matmul_f32->m, "matmul_f32_m", matmul_f32_f32_len, matmul_f32_f32_data, "main", 3, sizeof(vk_mat_mat_push_constants), m_wg_denoms, warptile_m, 1);
ggml_vk_create_pipeline(device, device->pipeline_matmul_f32->s, "matmul_f32_s", matmul_f32_f32_len, matmul_f32_f32_data, "main", 3, sizeof(vk_mat_mat_push_constants), s_wg_denoms, warptile_s, 1);
ggml_vk_create_pipeline(device, device->pipeline_matmul_f32->a_l, "matmul_f32_aligned_l", matmul_f32_f32_aligned_len, matmul_f32_f32_aligned_data, "main", 3, sizeof(vk_mat_mat_push_constants), l_wg_denoms, warptile_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_matmul_f32->a_m, "matmul_f32_aligned_m", matmul_f32_f32_aligned_len, matmul_f32_f32_aligned_data, "main", 3, sizeof(vk_mat_mat_push_constants), m_wg_denoms, warptile_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_matmul_f32->a_s, "matmul_f32_aligned_s", matmul_f32_f32_aligned_len, matmul_f32_f32_aligned_data, "main", 3, sizeof(vk_mat_mat_push_constants), s_wg_denoms, warptile_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_matmul_f32_f16->l, "matmul_f32_f16_l", matmul_f32_f16_len, matmul_f32_f16_data, "main", 3, sizeof(vk_mat_mat_push_constants), l_wg_denoms, warptile_l, 1);
ggml_vk_create_pipeline(device, device->pipeline_matmul_f32_f16->m, "matmul_f32_f16_m", matmul_f32_f16_len, matmul_f32_f16_data, "main", 3, sizeof(vk_mat_mat_push_constants), m_wg_denoms, warptile_m, 1);
ggml_vk_create_pipeline(device, device->pipeline_matmul_f32_f16->s, "matmul_f32_f16_s", matmul_f32_f16_len, matmul_f32_f16_data, "main", 3, sizeof(vk_mat_mat_push_constants), s_wg_denoms, warptile_s, 1);
ggml_vk_create_pipeline(device, device->pipeline_matmul_f32_f16->a_l, "matmul_f32_f16_aligned_l", matmul_f32_f16_aligned_len, matmul_f32_f16_aligned_data, "main", 3, sizeof(vk_mat_mat_push_constants), l_wg_denoms, warptile_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_matmul_f32_f16->a_m, "matmul_f32_f16_aligned_m", matmul_f32_f16_aligned_len, matmul_f32_f16_aligned_data, "main", 3, sizeof(vk_mat_mat_push_constants), m_wg_denoms, warptile_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_matmul_f32_f16->a_s, "matmul_f32_f16_aligned_s", matmul_f32_f16_aligned_len, matmul_f32_f16_aligned_data, "main", 3, sizeof(vk_mat_mat_push_constants), s_wg_denoms, warptile_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_matmul_f16->l, "matmul_f16_l", matmul_f16_len, matmul_f16_data, "main", 3, sizeof(vk_mat_mat_push_constants), l_wg_denoms, warptile_l, 1);
ggml_vk_create_pipeline(device, device->pipeline_matmul_f16->m, "matmul_f16_m", matmul_f16_len, matmul_f16_data, "main", 3, sizeof(vk_mat_mat_push_constants), m_wg_denoms, warptile_m, 1);
ggml_vk_create_pipeline(device, device->pipeline_matmul_f16->s, "matmul_f16_s", matmul_f16_len, matmul_f16_data, "main", 3, sizeof(vk_mat_mat_push_constants), s_wg_denoms, warptile_s, 1);
ggml_vk_create_pipeline(device, device->pipeline_matmul_f16->a_l, "matmul_f16_aligned_l", matmul_f16_aligned_len, matmul_f16_aligned_data, "main", 3, sizeof(vk_mat_mat_push_constants), l_wg_denoms, warptile_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_matmul_f16->a_m, "matmul_f16_aligned_m", matmul_f16_aligned_len, matmul_f16_aligned_data, "main", 3, sizeof(vk_mat_mat_push_constants), m_wg_denoms, warptile_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_matmul_f16->a_s, "matmul_f16_aligned_s", matmul_f16_aligned_len, matmul_f16_aligned_data, "main", 3, sizeof(vk_mat_mat_push_constants), s_wg_denoms, warptile_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_matmul_f16_f32->l, "matmul_f16_f32_l", matmul_f16_f32_len, matmul_f16_f32_data, "main", 3, sizeof(vk_mat_mat_push_constants), l_wg_denoms, warptile_l, 1);
ggml_vk_create_pipeline(device, device->pipeline_matmul_f16_f32->m, "matmul_f16_f32_m", matmul_f16_f32_len, matmul_f16_f32_data, "main", 3, sizeof(vk_mat_mat_push_constants), m_wg_denoms, warptile_m, 1);
ggml_vk_create_pipeline(device, device->pipeline_matmul_f16_f32->s, "matmul_f16_f32_s", matmul_f16_f32_len, matmul_f16_f32_data, "main", 3, sizeof(vk_mat_mat_push_constants), s_wg_denoms, warptile_s, 1);
ggml_vk_create_pipeline(device, device->pipeline_matmul_f16_f32->a_l, "matmul_f16_f32_aligned_l", matmul_f16_f32_aligned_len, matmul_f16_f32_aligned_data, "main", 3, sizeof(vk_mat_mat_push_constants), l_wg_denoms, warptile_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_matmul_f16_f32->a_m, "matmul_f16_f32_aligned_m", matmul_f16_f32_aligned_len, matmul_f16_f32_aligned_data, "main", 3, sizeof(vk_mat_mat_push_constants), m_wg_denoms, warptile_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_matmul_f16_f32->a_s, "matmul_f16_f32_aligned_s", matmul_f16_f32_aligned_len, matmul_f16_f32_aligned_data, "main", 3, sizeof(vk_mat_mat_push_constants), s_wg_denoms, warptile_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q4_0]->l, "matmul_q4_0_f32_l", matmul_q4_0_f32_len, matmul_q4_0_f32_data, "main", 3, sizeof(vk_mat_mat_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q4_0]->m, "matmul_q4_0_f32_m", matmul_q4_0_f32_len, matmul_q4_0_f32_data, "main", 3, sizeof(vk_mat_mat_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q4_0]->s, "matmul_q4_0_f32_s", matmul_q4_0_f32_len, matmul_q4_0_f32_data, "main", 3, sizeof(vk_mat_mat_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q4_0]->a_l, "matmul_q4_0_f32_aligned_l", matmul_q4_0_f32_aligned_len, matmul_q4_0_f32_aligned_data, "main", 3, sizeof(vk_mat_mat_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q4_0]->a_m, "matmul_q4_0_f32_aligned_m", matmul_q4_0_f32_aligned_len, matmul_q4_0_f32_aligned_data, "main", 3, sizeof(vk_mat_mat_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q4_0]->a_s, "matmul_q4_0_f32_aligned_s", matmul_q4_0_f32_aligned_len, matmul_q4_0_f32_aligned_data, "main", 3, sizeof(vk_mat_mat_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q4_1]->l, "matmul_q4_1_f32_l", matmul_q4_1_f32_len, matmul_q4_1_f32_data, "main", 3, sizeof(vk_mat_mat_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q4_1]->m, "matmul_q4_1_f32_m", matmul_q4_1_f32_len, matmul_q4_1_f32_data, "main", 3, sizeof(vk_mat_mat_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q4_1]->s, "matmul_q4_1_f32_s", matmul_q4_1_f32_len, matmul_q4_1_f32_data, "main", 3, sizeof(vk_mat_mat_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q4_1]->a_l, "matmul_q4_1_f32_aligned_l", matmul_q4_1_f32_aligned_len, matmul_q4_1_f32_aligned_data, "main", 3, sizeof(vk_mat_mat_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q4_1]->a_m, "matmul_q4_1_f32_aligned_m", matmul_q4_1_f32_aligned_len, matmul_q4_1_f32_aligned_data, "main", 3, sizeof(vk_mat_mat_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q4_1]->a_s, "matmul_q4_1_f32_aligned_s", matmul_q4_1_f32_aligned_len, matmul_q4_1_f32_aligned_data, "main", 3, sizeof(vk_mat_mat_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q5_0]->l, "matmul_q5_0_f32_l", matmul_q5_0_f32_len, matmul_q5_0_f32_data, "main", 3, sizeof(vk_mat_mat_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q5_0]->m, "matmul_q5_0_f32_m", matmul_q5_0_f32_len, matmul_q5_0_f32_data, "main", 3, sizeof(vk_mat_mat_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q5_0]->s, "matmul_q5_0_f32_s", matmul_q5_0_f32_len, matmul_q5_0_f32_data, "main", 3, sizeof(vk_mat_mat_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q5_0]->a_l, "matmul_q5_0_f32_aligned_l", matmul_q5_0_f32_aligned_len, matmul_q5_0_f32_aligned_data, "main", 3, sizeof(vk_mat_mat_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q5_0]->a_m, "matmul_q5_0_f32_aligned_m", matmul_q5_0_f32_aligned_len, matmul_q5_0_f32_aligned_data, "main", 3, sizeof(vk_mat_mat_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q5_0]->a_s, "matmul_q5_0_f32_aligned_s", matmul_q5_0_f32_aligned_len, matmul_q5_0_f32_aligned_data, "main", 3, sizeof(vk_mat_mat_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q5_1]->l, "matmul_q5_1_f32_l", matmul_q5_1_f32_len, matmul_q5_1_f32_data, "main", 3, sizeof(vk_mat_mat_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q5_1]->m, "matmul_q5_1_f32_m", matmul_q5_1_f32_len, matmul_q5_1_f32_data, "main", 3, sizeof(vk_mat_mat_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q5_1]->s, "matmul_q5_1_f32_s", matmul_q5_1_f32_len, matmul_q5_1_f32_data, "main", 3, sizeof(vk_mat_mat_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q5_1]->a_l, "matmul_q5_1_f32_aligned_l", matmul_q5_1_f32_aligned_len, matmul_q5_1_f32_aligned_data, "main", 3, sizeof(vk_mat_mat_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q5_1]->a_m, "matmul_q5_1_f32_aligned_m", matmul_q5_1_f32_aligned_len, matmul_q5_1_f32_aligned_data, "main", 3, sizeof(vk_mat_mat_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q5_1]->a_s, "matmul_q5_1_f32_aligned_s", matmul_q5_1_f32_aligned_len, matmul_q5_1_f32_aligned_data, "main", 3, sizeof(vk_mat_mat_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q8_0]->l, "matmul_q8_0_f32_l", matmul_q8_0_f32_len, matmul_q8_0_f32_data, "main", 3, sizeof(vk_mat_mat_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q8_0]->m, "matmul_q8_0_f32_m", matmul_q8_0_f32_len, matmul_q8_0_f32_data, "main", 3, sizeof(vk_mat_mat_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q8_0]->s, "matmul_q8_0_f32_s", matmul_q8_0_f32_len, matmul_q8_0_f32_data, "main", 3, sizeof(vk_mat_mat_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q8_0]->a_l, "matmul_q8_0_f32_aligned_l", matmul_q8_0_f32_aligned_len, matmul_q8_0_f32_aligned_data, "main", 3, sizeof(vk_mat_mat_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q8_0]->a_m, "matmul_q8_0_f32_aligned_m", matmul_q8_0_f32_aligned_len, matmul_q8_0_f32_aligned_data, "main", 3, sizeof(vk_mat_mat_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q8_0]->a_s, "matmul_q8_0_f32_aligned_s", matmul_q8_0_f32_aligned_len, matmul_q8_0_f32_aligned_data, "main", 3, sizeof(vk_mat_mat_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q2_K]->l, "matmul_q2_k_f32_l", matmul_q2_k_f32_len, matmul_q2_k_f32_data, "main", 3, sizeof(vk_mat_mat_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q2_K]->m, "matmul_q2_k_f32_m", matmul_q2_k_f32_len, matmul_q2_k_f32_data, "main", 3, sizeof(vk_mat_mat_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q2_K]->s, "matmul_q2_k_f32_s", matmul_q2_k_f32_len, matmul_q2_k_f32_data, "main", 3, sizeof(vk_mat_mat_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q2_K]->a_l, "matmul_q2_k_f32_aligned_l", matmul_q2_k_f32_aligned_len, matmul_q2_k_f32_aligned_data, "main", 3, sizeof(vk_mat_mat_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q2_K]->a_m, "matmul_q2_k_f32_aligned_m", matmul_q2_k_f32_aligned_len, matmul_q2_k_f32_aligned_data, "main", 3, sizeof(vk_mat_mat_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q2_K]->a_s, "matmul_q2_k_f32_aligned_s", matmul_q2_k_f32_aligned_len, matmul_q2_k_f32_aligned_data, "main", 3, sizeof(vk_mat_mat_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q3_K]->l, "matmul_q3_k_f32_l", matmul_q3_k_f32_len, matmul_q3_k_f32_data, "main", 3, sizeof(vk_mat_mat_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q3_K]->m, "matmul_q3_k_f32_m", matmul_q3_k_f32_len, matmul_q3_k_f32_data, "main", 3, sizeof(vk_mat_mat_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q3_K]->s, "matmul_q3_k_f32_s", matmul_q3_k_f32_len, matmul_q3_k_f32_data, "main", 3, sizeof(vk_mat_mat_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q3_K]->a_l, "matmul_q3_k_f32_aligned_l", matmul_q3_k_f32_aligned_len, matmul_q3_k_f32_aligned_data, "main", 3, sizeof(vk_mat_mat_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q3_K]->a_m, "matmul_q3_k_f32_aligned_m", matmul_q3_k_f32_aligned_len, matmul_q3_k_f32_aligned_data, "main", 3, sizeof(vk_mat_mat_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q3_K]->a_s, "matmul_q3_k_f32_aligned_s", matmul_q3_k_f32_aligned_len, matmul_q3_k_f32_aligned_data, "main", 3, sizeof(vk_mat_mat_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q4_K]->l, "matmul_q4_k_f32_l", matmul_q4_k_f32_len, matmul_q4_k_f32_data, "main", 3, sizeof(vk_mat_mat_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q4_K]->m, "matmul_q4_k_f32_m", matmul_q4_k_f32_len, matmul_q4_k_f32_data, "main", 3, sizeof(vk_mat_mat_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q4_K]->s, "matmul_q4_k_f32_s", matmul_q4_k_f32_len, matmul_q4_k_f32_data, "main", 3, sizeof(vk_mat_mat_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q4_K]->a_l, "matmul_q4_k_f32_aligned_l", matmul_q4_k_f32_aligned_len, matmul_q4_k_f32_aligned_data, "main", 3, sizeof(vk_mat_mat_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q4_K]->a_m, "matmul_q4_k_f32_aligned_m", matmul_q4_k_f32_aligned_len, matmul_q4_k_f32_aligned_data, "main", 3, sizeof(vk_mat_mat_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q4_K]->a_s, "matmul_q4_k_f32_aligned_s", matmul_q4_k_f32_aligned_len, matmul_q4_k_f32_aligned_data, "main", 3, sizeof(vk_mat_mat_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q5_K]->l, "matmul_q5_k_f32_l", matmul_q5_k_f32_len, matmul_q5_k_f32_data, "main", 3, sizeof(vk_mat_mat_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q5_K]->m, "matmul_q5_k_f32_m", matmul_q5_k_f32_len, matmul_q5_k_f32_data, "main", 3, sizeof(vk_mat_mat_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q5_K]->s, "matmul_q5_k_f32_s", matmul_q5_k_f32_len, matmul_q5_k_f32_data, "main", 3, sizeof(vk_mat_mat_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q5_K]->a_l, "matmul_q5_k_f32_aligned_l", matmul_q5_k_f32_aligned_len, matmul_q5_k_f32_aligned_data, "main", 3, sizeof(vk_mat_mat_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q5_K]->a_m, "matmul_q5_k_f32_aligned_m", matmul_q5_k_f32_aligned_len, matmul_q5_k_f32_aligned_data, "main", 3, sizeof(vk_mat_mat_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q5_K]->a_s, "matmul_q5_k_f32_aligned_s", matmul_q5_k_f32_aligned_len, matmul_q5_k_f32_aligned_data, "main", 3, sizeof(vk_mat_mat_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q6_K]->l, "matmul_q6_k_f32_l", matmul_q6_k_f32_len, matmul_q6_k_f32_data, "main", 3, sizeof(vk_mat_mat_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q6_K]->m, "matmul_q6_k_f32_m", matmul_q6_k_f32_len, matmul_q6_k_f32_data, "main", 3, sizeof(vk_mat_mat_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q6_K]->s, "matmul_q6_k_f32_s", matmul_q6_k_f32_len, matmul_q6_k_f32_data, "main", 3, sizeof(vk_mat_mat_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q6_K]->a_l, "matmul_q6_k_f32_aligned_l", matmul_q6_k_f32_aligned_len, matmul_q6_k_f32_aligned_data, "main", 3, sizeof(vk_mat_mat_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q6_K]->a_m, "matmul_q6_k_f32_aligned_m", matmul_q6_k_f32_aligned_len, matmul_q6_k_f32_aligned_data, "main", 3, sizeof(vk_mat_mat_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q6_K]->a_s, "matmul_q6_k_f32_aligned_s", matmul_q6_k_f32_aligned_len, matmul_q6_k_f32_aligned_data, "main", 3, sizeof(vk_mat_mat_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_matmul_id_f32->l, "matmul_id_f32_l", matmul_id_f32_f32_len, matmul_id_f32_f32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), l_wg_denoms, warptile_l, 1);
ggml_vk_create_pipeline(device, device->pipeline_matmul_id_f32->m, "matmul_id_f32_m", matmul_id_f32_f32_len, matmul_id_f32_f32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), m_wg_denoms, warptile_m, 1);
ggml_vk_create_pipeline(device, device->pipeline_matmul_id_f32->s, "matmul_id_f32_s", matmul_id_f32_f32_len, matmul_id_f32_f32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), s_wg_denoms, warptile_s, 1);
ggml_vk_create_pipeline(device, device->pipeline_matmul_id_f32->a_l, "matmul_id_f32_aligned_l", matmul_id_f32_f32_aligned_len, matmul_id_f32_f32_aligned_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), l_wg_denoms, warptile_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_matmul_id_f32->a_m, "matmul_id_f32_aligned_m", matmul_id_f32_f32_aligned_len, matmul_id_f32_f32_aligned_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), m_wg_denoms, warptile_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_matmul_id_f32->a_s, "matmul_id_f32_aligned_s", matmul_id_f32_f32_aligned_len, matmul_id_f32_f32_aligned_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), s_wg_denoms, warptile_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_matmul_id_f16->l, "matmul_id_f16_l", matmul_id_f16_len, matmul_id_f16_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), l_wg_denoms, warptile_l, 1);
ggml_vk_create_pipeline(device, device->pipeline_matmul_id_f16->m, "matmul_id_f16_m", matmul_id_f16_len, matmul_id_f16_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), m_wg_denoms, warptile_m, 1);
ggml_vk_create_pipeline(device, device->pipeline_matmul_id_f16->s, "matmul_id_f16_s", matmul_id_f16_len, matmul_id_f16_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), s_wg_denoms, warptile_s, 1);
ggml_vk_create_pipeline(device, device->pipeline_matmul_id_f16->a_l, "matmul_id_f16_aligned_l", matmul_id_f16_aligned_len, matmul_id_f16_aligned_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), l_wg_denoms, warptile_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_matmul_id_f16->a_m, "matmul_id_f16_aligned_m", matmul_id_f16_aligned_len, matmul_id_f16_aligned_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), m_wg_denoms, warptile_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_matmul_id_f16->a_s, "matmul_id_f16_aligned_s", matmul_id_f16_aligned_len, matmul_id_f16_aligned_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), s_wg_denoms, warptile_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_matmul_id_f16_f32->l, "matmul_id_f16_f32_l", matmul_id_f16_f32_len, matmul_id_f16_f32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), l_wg_denoms, warptile_l, 1);
ggml_vk_create_pipeline(device, device->pipeline_matmul_id_f16_f32->m, "matmul_id_f16_f32_m", matmul_id_f16_f32_len, matmul_id_f16_f32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), m_wg_denoms, warptile_m, 1);
ggml_vk_create_pipeline(device, device->pipeline_matmul_id_f16_f32->s, "matmul_id_f16_f32_s", matmul_id_f16_f32_len, matmul_id_f16_f32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), s_wg_denoms, warptile_s, 1);
ggml_vk_create_pipeline(device, device->pipeline_matmul_id_f16_f32->a_l, "matmul_id_f16_f32_aligned_l", matmul_id_f16_f32_aligned_len, matmul_id_f16_f32_aligned_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), l_wg_denoms, warptile_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_matmul_id_f16_f32->a_m, "matmul_id_f16_f32_aligned_m", matmul_id_f16_f32_aligned_len, matmul_id_f16_f32_aligned_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), m_wg_denoms, warptile_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_matmul_id_f16_f32->a_s, "matmul_id_f16_f32_aligned_s", matmul_id_f16_f32_aligned_len, matmul_id_f16_f32_aligned_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), s_wg_denoms, warptile_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q4_0]->l, "matmul_id_q4_0_f32_l", matmul_id_q4_0_f32_len, matmul_id_q4_0_f32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q4_0]->m, "matmul_id_q4_0_f32_m", matmul_id_q4_0_f32_len, matmul_id_q4_0_f32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q4_0]->s, "matmul_id_q4_0_f32_s", matmul_id_q4_0_f32_len, matmul_id_q4_0_f32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q4_0]->a_l, "matmul_id_q4_0_f32_aligned_l", matmul_id_q4_0_f32_aligned_len, matmul_id_q4_0_f32_aligned_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q4_0]->a_m, "matmul_id_q4_0_f32_aligned_m", matmul_id_q4_0_f32_aligned_len, matmul_id_q4_0_f32_aligned_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q4_0]->a_s, "matmul_id_q4_0_f32_aligned_s", matmul_id_q4_0_f32_aligned_len, matmul_id_q4_0_f32_aligned_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q4_1]->l, "matmul_id_q4_1_f32_l", matmul_id_q4_1_f32_len, matmul_id_q4_1_f32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q4_1]->m, "matmul_id_q4_1_f32_m", matmul_id_q4_1_f32_len, matmul_id_q4_1_f32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q4_1]->s, "matmul_id_q4_1_f32_s", matmul_id_q4_1_f32_len, matmul_id_q4_1_f32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q4_1]->a_l, "matmul_id_q4_1_f32_aligned_l", matmul_id_q4_1_f32_aligned_len, matmul_id_q4_1_f32_aligned_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q4_1]->a_m, "matmul_id_q4_1_f32_aligned_m", matmul_id_q4_1_f32_aligned_len, matmul_id_q4_1_f32_aligned_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q4_1]->a_s, "matmul_id_q4_1_f32_aligned_s", matmul_id_q4_1_f32_aligned_len, matmul_id_q4_1_f32_aligned_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q5_0]->l, "matmul_id_q5_0_f32_l", matmul_id_q5_0_f32_len, matmul_id_q5_0_f32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q5_0]->m, "matmul_id_q5_0_f32_m", matmul_id_q5_0_f32_len, matmul_id_q5_0_f32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q5_0]->s, "matmul_id_q5_0_f32_s", matmul_id_q5_0_f32_len, matmul_id_q5_0_f32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q5_0]->a_l, "matmul_id_q5_0_f32_aligned_l", matmul_id_q5_0_f32_aligned_len, matmul_id_q5_0_f32_aligned_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q5_0]->a_m, "matmul_id_q5_0_f32_aligned_m", matmul_id_q5_0_f32_aligned_len, matmul_id_q5_0_f32_aligned_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q5_0]->a_s, "matmul_id_q5_0_f32_aligned_s", matmul_id_q5_0_f32_aligned_len, matmul_id_q5_0_f32_aligned_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q5_1]->l, "matmul_id_q5_1_f32_l", matmul_id_q5_1_f32_len, matmul_id_q5_1_f32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q5_1]->m, "matmul_id_q5_1_f32_m", matmul_id_q5_1_f32_len, matmul_id_q5_1_f32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q5_1]->s, "matmul_id_q5_1_f32_s", matmul_id_q5_1_f32_len, matmul_id_q5_1_f32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q5_1]->a_l, "matmul_id_q5_1_f32_aligned_l", matmul_id_q5_1_f32_aligned_len, matmul_id_q5_1_f32_aligned_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q5_1]->a_m, "matmul_id_q5_1_f32_aligned_m", matmul_id_q5_1_f32_aligned_len, matmul_id_q5_1_f32_aligned_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q5_1]->a_s, "matmul_id_q5_1_f32_aligned_s", matmul_id_q5_1_f32_aligned_len, matmul_id_q5_1_f32_aligned_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q8_0]->l, "matmul_id_q8_0_f32_l", matmul_id_q8_0_f32_len, matmul_id_q8_0_f32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q8_0]->m, "matmul_id_q8_0_f32_m", matmul_id_q8_0_f32_len, matmul_id_q8_0_f32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q8_0]->s, "matmul_id_q8_0_f32_s", matmul_id_q8_0_f32_len, matmul_id_q8_0_f32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q8_0]->a_l, "matmul_id_q8_0_f32_aligned_l", matmul_id_q8_0_f32_aligned_len, matmul_id_q8_0_f32_aligned_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q8_0]->a_m, "matmul_id_q8_0_f32_aligned_m", matmul_id_q8_0_f32_aligned_len, matmul_id_q8_0_f32_aligned_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q8_0]->a_s, "matmul_id_q8_0_f32_aligned_s", matmul_id_q8_0_f32_aligned_len, matmul_id_q8_0_f32_aligned_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q2_K]->l, "matmul_id_q2_k_f32_l", matmul_id_q2_k_f32_len, matmul_id_q2_k_f32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q2_K]->m, "matmul_id_q2_k_f32_m", matmul_id_q2_k_f32_len, matmul_id_q2_k_f32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q2_K]->s, "matmul_id_q2_k_f32_s", matmul_id_q2_k_f32_len, matmul_id_q2_k_f32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q2_K]->a_l, "matmul_id_q2_k_f32_aligned_l", matmul_id_q2_k_f32_aligned_len, matmul_id_q2_k_f32_aligned_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q2_K]->a_m, "matmul_id_q2_k_f32_aligned_m", matmul_id_q2_k_f32_aligned_len, matmul_id_q2_k_f32_aligned_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q2_K]->a_s, "matmul_id_q2_k_f32_aligned_s", matmul_id_q2_k_f32_aligned_len, matmul_id_q2_k_f32_aligned_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q3_K]->l, "matmul_id_q3_k_f32_l", matmul_id_q3_k_f32_len, matmul_id_q3_k_f32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q3_K]->m, "matmul_id_q3_k_f32_m", matmul_id_q3_k_f32_len, matmul_id_q3_k_f32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q3_K]->s, "matmul_id_q3_k_f32_s", matmul_id_q3_k_f32_len, matmul_id_q3_k_f32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q3_K]->a_l, "matmul_id_q3_k_f32_aligned_l", matmul_id_q3_k_f32_aligned_len, matmul_id_q3_k_f32_aligned_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q3_K]->a_m, "matmul_id_q3_k_f32_aligned_m", matmul_id_q3_k_f32_aligned_len, matmul_id_q3_k_f32_aligned_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q3_K]->a_s, "matmul_id_q3_k_f32_aligned_s", matmul_id_q3_k_f32_aligned_len, matmul_id_q3_k_f32_aligned_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q4_K]->l, "matmul_id_q4_k_f32_l", matmul_id_q4_k_f32_len, matmul_id_q4_k_f32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q4_K]->m, "matmul_id_q4_k_f32_m", matmul_id_q4_k_f32_len, matmul_id_q4_k_f32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q4_K]->s, "matmul_id_q4_k_f32_s", matmul_id_q4_k_f32_len, matmul_id_q4_k_f32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q4_K]->a_l, "matmul_id_q4_k_f32_aligned_l", matmul_id_q4_k_f32_aligned_len, matmul_id_q4_k_f32_aligned_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q4_K]->a_m, "matmul_id_q4_k_f32_aligned_m", matmul_id_q4_k_f32_aligned_len, matmul_id_q4_k_f32_aligned_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q4_K]->a_s, "matmul_id_q4_k_f32_aligned_s", matmul_id_q4_k_f32_aligned_len, matmul_id_q4_k_f32_aligned_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q5_K]->l, "matmul_id_q5_k_f32_l", matmul_id_q5_k_f32_len, matmul_id_q5_k_f32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q5_K]->m, "matmul_id_q5_k_f32_m", matmul_id_q5_k_f32_len, matmul_id_q5_k_f32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q5_K]->s, "matmul_id_q5_k_f32_s", matmul_id_q5_k_f32_len, matmul_id_q5_k_f32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q5_K]->a_l, "matmul_id_q5_k_f32_aligned_l", matmul_id_q5_k_f32_aligned_len, matmul_id_q5_k_f32_aligned_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q5_K]->a_m, "matmul_id_q5_k_f32_aligned_m", matmul_id_q5_k_f32_aligned_len, matmul_id_q5_k_f32_aligned_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q5_K]->a_s, "matmul_id_q5_k_f32_aligned_s", matmul_id_q5_k_f32_aligned_len, matmul_id_q5_k_f32_aligned_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q6_K]->l, "matmul_id_q6_k_f32_l", matmul_id_q6_k_f32_len, matmul_id_q6_k_f32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q6_K]->m, "matmul_id_q6_k_f32_m", matmul_id_q6_k_f32_len, matmul_id_q6_k_f32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q6_K]->s, "matmul_id_q6_k_f32_s", matmul_id_q6_k_f32_len, matmul_id_q6_k_f32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q6_K]->a_l, "matmul_id_q6_k_f32_aligned_l", matmul_id_q6_k_f32_aligned_len, matmul_id_q6_k_f32_aligned_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q6_K]->a_m, "matmul_id_q6_k_f32_aligned_m", matmul_id_q6_k_f32_aligned_len, matmul_id_q6_k_f32_aligned_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q6_K]->a_s, "matmul_id_q6_k_f32_aligned_s", matmul_id_q6_k_f32_aligned_len, matmul_id_q6_k_f32_aligned_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
} else {
ggml_vk_create_pipeline(device, device->pipeline_matmul_f32->l, "matmul_f32_l", matmul_f32_f32_fp32_len, matmul_f32_f32_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), l_wg_denoms, warptile_l, 1);
ggml_vk_create_pipeline(device, device->pipeline_matmul_f32->m, "matmul_f32_m", matmul_f32_f32_fp32_len, matmul_f32_f32_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), m_wg_denoms, warptile_m, 1);
ggml_vk_create_pipeline(device, device->pipeline_matmul_f32->s, "matmul_f32_s", matmul_f32_f32_fp32_len, matmul_f32_f32_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), s_wg_denoms, warptile_s, 1);
ggml_vk_create_pipeline(device, device->pipeline_matmul_f32->a_l, "matmul_f32_aligned_l", matmul_f32_f32_aligned_fp32_len, matmul_f32_f32_aligned_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), l_wg_denoms, warptile_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_matmul_f32->a_m, "matmul_f32_aligned_m", matmul_f32_f32_aligned_fp32_len, matmul_f32_f32_aligned_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), m_wg_denoms, warptile_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_matmul_f32->a_s, "matmul_f32_aligned_s", matmul_f32_f32_aligned_fp32_len, matmul_f32_f32_aligned_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), s_wg_denoms, warptile_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_matmul_f32_f16->l, "matmul_f32_f16_l", matmul_f32_f16_fp32_len, matmul_f32_f16_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), l_wg_denoms, warptile_l, 1);
ggml_vk_create_pipeline(device, device->pipeline_matmul_f32_f16->m, "matmul_f32_f16_m", matmul_f32_f16_fp32_len, matmul_f32_f16_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), m_wg_denoms, warptile_m, 1);
ggml_vk_create_pipeline(device, device->pipeline_matmul_f32_f16->s, "matmul_f32_f16_s", matmul_f32_f16_fp32_len, matmul_f32_f16_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), s_wg_denoms, warptile_s, 1);
ggml_vk_create_pipeline(device, device->pipeline_matmul_f32_f16->a_l, "matmul_f32_f16_aligned_l", matmul_f32_f16_aligned_fp32_len, matmul_f32_f16_aligned_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), l_wg_denoms, warptile_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_matmul_f32_f16->a_m, "matmul_f32_f16_aligned_m", matmul_f32_f16_aligned_fp32_len, matmul_f32_f16_aligned_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), m_wg_denoms, warptile_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_matmul_f32_f16->a_s, "matmul_f32_f16_aligned_s", matmul_f32_f16_aligned_fp32_len, matmul_f32_f16_aligned_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), s_wg_denoms, warptile_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_matmul_f16->l, "matmul_f16_l", matmul_f16_fp32_len, matmul_f16_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), l_wg_denoms, warptile_l, 1);
ggml_vk_create_pipeline(device, device->pipeline_matmul_f16->m, "matmul_f16_m", matmul_f16_fp32_len, matmul_f16_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), m_wg_denoms, warptile_m, 1);
ggml_vk_create_pipeline(device, device->pipeline_matmul_f16->s, "matmul_f16_s", matmul_f16_fp32_len, matmul_f16_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), s_wg_denoms, warptile_s, 1);
ggml_vk_create_pipeline(device, device->pipeline_matmul_f16->a_l, "matmul_f16_aligned_l", matmul_f16_aligned_fp32_len, matmul_f16_aligned_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), l_wg_denoms, warptile_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_matmul_f16->a_m, "matmul_f16_aligned_m", matmul_f16_aligned_fp32_len, matmul_f16_aligned_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), m_wg_denoms, warptile_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_matmul_f16->a_s, "matmul_f16_aligned_s", matmul_f16_aligned_fp32_len, matmul_f16_aligned_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), s_wg_denoms, warptile_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_matmul_f16_f32->l, "matmul_f16_f32_l", matmul_f16_f32_fp32_len, matmul_f16_f32_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), l_wg_denoms, warptile_l, 1);
ggml_vk_create_pipeline(device, device->pipeline_matmul_f16_f32->m, "matmul_f16_f32_m", matmul_f16_f32_fp32_len, matmul_f16_f32_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), m_wg_denoms, warptile_m, 1);
ggml_vk_create_pipeline(device, device->pipeline_matmul_f16_f32->s, "matmul_f16_f32_s", matmul_f16_f32_fp32_len, matmul_f16_f32_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), s_wg_denoms, warptile_s, 1);
ggml_vk_create_pipeline(device, device->pipeline_matmul_f16_f32->a_l, "matmul_f16_f32_aligned_l", matmul_f16_f32_aligned_fp32_len, matmul_f16_f32_aligned_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), l_wg_denoms, warptile_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_matmul_f16_f32->a_m, "matmul_f16_f32_aligned_m", matmul_f16_f32_aligned_fp32_len, matmul_f16_f32_aligned_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), m_wg_denoms, warptile_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_matmul_f16_f32->a_s, "matmul_f16_f32_aligned_s", matmul_f16_f32_aligned_fp32_len, matmul_f16_f32_aligned_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), s_wg_denoms, warptile_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q4_0]->l, "matmul_q4_0_f32_l", matmul_q4_0_f32_fp32_len, matmul_q4_0_f32_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q4_0]->m, "matmul_q4_0_f32_m", matmul_q4_0_f32_fp32_len, matmul_q4_0_f32_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q4_0]->s, "matmul_q4_0_f32_s", matmul_q4_0_f32_fp32_len, matmul_q4_0_f32_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q4_0]->a_l, "matmul_q4_0_f32_aligned_l", matmul_q4_0_f32_aligned_fp32_len, matmul_q4_0_f32_aligned_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q4_0]->a_m, "matmul_q4_0_f32_aligned_m", matmul_q4_0_f32_aligned_fp32_len, matmul_q4_0_f32_aligned_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q4_0]->a_s, "matmul_q4_0_f32_aligned_s", matmul_q4_0_f32_aligned_fp32_len, matmul_q4_0_f32_aligned_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q4_1]->l, "matmul_q4_1_f32_l", matmul_q4_1_f32_fp32_len, matmul_q4_1_f32_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q4_1]->m, "matmul_q4_1_f32_m", matmul_q4_1_f32_fp32_len, matmul_q4_1_f32_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q4_1]->s, "matmul_q4_1_f32_s", matmul_q4_1_f32_fp32_len, matmul_q4_1_f32_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q4_1]->a_l, "matmul_q4_1_f32_aligned_l", matmul_q4_1_f32_aligned_fp32_len, matmul_q4_1_f32_aligned_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q4_1]->a_m, "matmul_q4_1_f32_aligned_m", matmul_q4_1_f32_aligned_fp32_len, matmul_q4_1_f32_aligned_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q4_1]->a_s, "matmul_q4_1_f32_aligned_s", matmul_q4_1_f32_aligned_fp32_len, matmul_q4_1_f32_aligned_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q5_0]->l, "matmul_q5_0_f32_l", matmul_q5_0_f32_fp32_len, matmul_q5_0_f32_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q5_0]->m, "matmul_q5_0_f32_m", matmul_q5_0_f32_fp32_len, matmul_q5_0_f32_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q5_0]->s, "matmul_q5_0_f32_s", matmul_q5_0_f32_fp32_len, matmul_q5_0_f32_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q5_0]->a_l, "matmul_q5_0_f32_aligned_l", matmul_q5_0_f32_aligned_fp32_len, matmul_q5_0_f32_aligned_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q5_0]->a_m, "matmul_q5_0_f32_aligned_m", matmul_q5_0_f32_aligned_fp32_len, matmul_q5_0_f32_aligned_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q5_0]->a_s, "matmul_q5_0_f32_aligned_s", matmul_q5_0_f32_aligned_fp32_len, matmul_q5_0_f32_aligned_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q5_1]->l, "matmul_q5_1_f32_l", matmul_q5_1_f32_fp32_len, matmul_q5_1_f32_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q5_1]->m, "matmul_q5_1_f32_m", matmul_q5_1_f32_fp32_len, matmul_q5_1_f32_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q5_1]->s, "matmul_q5_1_f32_s", matmul_q5_1_f32_fp32_len, matmul_q5_1_f32_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q5_1]->a_l, "matmul_q5_1_f32_aligned_l", matmul_q5_1_f32_aligned_fp32_len, matmul_q5_1_f32_aligned_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q5_1]->a_m, "matmul_q5_1_f32_aligned_m", matmul_q5_1_f32_aligned_fp32_len, matmul_q5_1_f32_aligned_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q5_1]->a_s, "matmul_q5_1_f32_aligned_s", matmul_q5_1_f32_aligned_fp32_len, matmul_q5_1_f32_aligned_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q8_0]->l, "matmul_q8_0_f32_l", matmul_q8_0_f32_fp32_len, matmul_q8_0_f32_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q8_0]->m, "matmul_q8_0_f32_m", matmul_q8_0_f32_fp32_len, matmul_q8_0_f32_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q8_0]->s, "matmul_q8_0_f32_s", matmul_q8_0_f32_fp32_len, matmul_q8_0_f32_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q8_0]->a_l, "matmul_q8_0_f32_aligned_l", matmul_q8_0_f32_aligned_fp32_len, matmul_q8_0_f32_aligned_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q8_0]->a_m, "matmul_q8_0_f32_aligned_m", matmul_q8_0_f32_aligned_fp32_len, matmul_q8_0_f32_aligned_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q8_0]->a_s, "matmul_q8_0_f32_aligned_s", matmul_q8_0_f32_aligned_fp32_len, matmul_q8_0_f32_aligned_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q2_K]->l, "matmul_q2_k_f32_l", matmul_q2_k_f32_fp32_len, matmul_q2_k_f32_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q2_K]->m, "matmul_q2_k_f32_m", matmul_q2_k_f32_fp32_len, matmul_q2_k_f32_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q2_K]->s, "matmul_q2_k_f32_s", matmul_q2_k_f32_fp32_len, matmul_q2_k_f32_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q2_K]->a_l, "matmul_q2_k_f32_aligned_l", matmul_q2_k_f32_aligned_fp32_len, matmul_q2_k_f32_aligned_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q2_K]->a_m, "matmul_q2_k_f32_aligned_m", matmul_q2_k_f32_aligned_fp32_len, matmul_q2_k_f32_aligned_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q2_K]->a_s, "matmul_q2_k_f32_aligned_s", matmul_q2_k_f32_aligned_fp32_len, matmul_q2_k_f32_aligned_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q3_K]->l, "matmul_q3_k_f32_l", matmul_q3_k_f32_fp32_len, matmul_q3_k_f32_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q3_K]->m, "matmul_q3_k_f32_m", matmul_q3_k_f32_fp32_len, matmul_q3_k_f32_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q3_K]->s, "matmul_q3_k_f32_s", matmul_q3_k_f32_fp32_len, matmul_q3_k_f32_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q3_K]->a_l, "matmul_q3_k_f32_aligned_l", matmul_q3_k_f32_aligned_fp32_len, matmul_q3_k_f32_aligned_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q3_K]->a_m, "matmul_q3_k_f32_aligned_m", matmul_q3_k_f32_aligned_fp32_len, matmul_q3_k_f32_aligned_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q3_K]->a_s, "matmul_q3_k_f32_aligned_s", matmul_q3_k_f32_aligned_fp32_len, matmul_q3_k_f32_aligned_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q4_K]->l, "matmul_q4_k_f32_l", matmul_q4_k_f32_fp32_len, matmul_q4_k_f32_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q4_K]->m, "matmul_q4_k_f32_m", matmul_q4_k_f32_fp32_len, matmul_q4_k_f32_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q4_K]->s, "matmul_q4_k_f32_s", matmul_q4_k_f32_fp32_len, matmul_q4_k_f32_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q4_K]->a_l, "matmul_q4_k_f32_aligned_l", matmul_q4_k_f32_aligned_fp32_len, matmul_q4_k_f32_aligned_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q4_K]->a_m, "matmul_q4_k_f32_aligned_m", matmul_q4_k_f32_aligned_fp32_len, matmul_q4_k_f32_aligned_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q4_K]->a_s, "matmul_q4_k_f32_aligned_s", matmul_q4_k_f32_aligned_fp32_len, matmul_q4_k_f32_aligned_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q5_K]->l, "matmul_q5_k_f32_l", matmul_q5_k_f32_fp32_len, matmul_q5_k_f32_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q5_K]->m, "matmul_q5_k_f32_m", matmul_q5_k_f32_fp32_len, matmul_q5_k_f32_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q5_K]->s, "matmul_q5_k_f32_s", matmul_q5_k_f32_fp32_len, matmul_q5_k_f32_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q5_K]->a_l, "matmul_q5_k_f32_aligned_l", matmul_q5_k_f32_aligned_fp32_len, matmul_q5_k_f32_aligned_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q5_K]->a_m, "matmul_q5_k_f32_aligned_m", matmul_q5_k_f32_aligned_fp32_len, matmul_q5_k_f32_aligned_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q5_K]->a_s, "matmul_q5_k_f32_aligned_s", matmul_q5_k_f32_aligned_fp32_len, matmul_q5_k_f32_aligned_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q6_K]->l, "matmul_q6_k_f32_l", matmul_q6_k_f32_fp32_len, matmul_q6_k_f32_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q6_K]->m, "matmul_q6_k_f32_m", matmul_q6_k_f32_fp32_len, matmul_q6_k_f32_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q6_K]->s, "matmul_q6_k_f32_s", matmul_q6_k_f32_fp32_len, matmul_q6_k_f32_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q6_K]->a_l, "matmul_q6_k_f32_aligned_l", matmul_q6_k_f32_aligned_fp32_len, matmul_q6_k_f32_aligned_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q6_K]->a_m, "matmul_q6_k_f32_aligned_m", matmul_q6_k_f32_aligned_fp32_len, matmul_q6_k_f32_aligned_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat[GGML_TYPE_Q6_K]->a_s, "matmul_q6_k_f32_aligned_s", matmul_q6_k_f32_aligned_fp32_len, matmul_q6_k_f32_aligned_fp32_data, "main", 3, sizeof(vk_mat_mat_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_matmul_id_f32->l, "matmul_id_f32_l", matmul_id_f32_f32_fp32_len, matmul_id_f32_f32_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), l_wg_denoms, warptile_l, 1);
ggml_vk_create_pipeline(device, device->pipeline_matmul_id_f32->m, "matmul_id_f32_m", matmul_id_f32_f32_fp32_len, matmul_id_f32_f32_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), m_wg_denoms, warptile_m, 1);
ggml_vk_create_pipeline(device, device->pipeline_matmul_id_f32->s, "matmul_id_f32_s", matmul_id_f32_f32_fp32_len, matmul_id_f32_f32_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), s_wg_denoms, warptile_s, 1);
ggml_vk_create_pipeline(device, device->pipeline_matmul_id_f32->a_l, "matmul_id_f32_aligned_l", matmul_id_f32_f32_aligned_fp32_len, matmul_id_f32_f32_aligned_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), l_wg_denoms, warptile_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_matmul_id_f32->a_m, "matmul_id_f32_aligned_m", matmul_id_f32_f32_aligned_fp32_len, matmul_id_f32_f32_aligned_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), m_wg_denoms, warptile_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_matmul_id_f32->a_s, "matmul_id_f32_aligned_s", matmul_id_f32_f32_aligned_fp32_len, matmul_id_f32_f32_aligned_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), s_wg_denoms, warptile_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_matmul_id_f16->l, "matmul_id_f16_l", matmul_id_f16_fp32_len, matmul_id_f16_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), l_wg_denoms, warptile_l, 1);
ggml_vk_create_pipeline(device, device->pipeline_matmul_id_f16->m, "matmul_id_f16_m", matmul_id_f16_fp32_len, matmul_id_f16_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), m_wg_denoms, warptile_m, 1);
ggml_vk_create_pipeline(device, device->pipeline_matmul_id_f16->s, "matmul_id_f16_s", matmul_id_f16_fp32_len, matmul_id_f16_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), s_wg_denoms, warptile_s, 1);
ggml_vk_create_pipeline(device, device->pipeline_matmul_id_f16->a_l, "matmul_id_f16_aligned_l", matmul_id_f16_aligned_fp32_len, matmul_id_f16_aligned_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), l_wg_denoms, warptile_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_matmul_id_f16->a_m, "matmul_id_f16_aligned_m", matmul_id_f16_aligned_fp32_len, matmul_id_f16_aligned_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), m_wg_denoms, warptile_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_matmul_id_f16->a_s, "matmul_id_f16_aligned_s", matmul_id_f16_aligned_fp32_len, matmul_id_f16_aligned_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), s_wg_denoms, warptile_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_matmul_id_f16_f32->l, "matmul_id_f16_f32_l", matmul_id_f16_f32_fp32_len, matmul_id_f16_f32_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), l_wg_denoms, warptile_l, 1);
ggml_vk_create_pipeline(device, device->pipeline_matmul_id_f16_f32->m, "matmul_id_f16_f32_m", matmul_id_f16_f32_fp32_len, matmul_id_f16_f32_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), m_wg_denoms, warptile_m, 1);
ggml_vk_create_pipeline(device, device->pipeline_matmul_id_f16_f32->s, "matmul_id_f16_f32_s", matmul_id_f16_f32_fp32_len, matmul_id_f16_f32_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), s_wg_denoms, warptile_s, 1);
ggml_vk_create_pipeline(device, device->pipeline_matmul_id_f16_f32->a_l, "matmul_id_f16_f32_aligned_l", matmul_id_f16_f32_aligned_fp32_len, matmul_id_f16_f32_aligned_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), l_wg_denoms, warptile_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_matmul_id_f16_f32->a_m, "matmul_id_f16_f32_aligned_m", matmul_id_f16_f32_aligned_fp32_len, matmul_id_f16_f32_aligned_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), m_wg_denoms, warptile_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_matmul_id_f16_f32->a_s, "matmul_id_f16_f32_aligned_s", matmul_id_f16_f32_aligned_fp32_len, matmul_id_f16_f32_aligned_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), s_wg_denoms, warptile_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q4_0]->l, "matmul_id_q4_0_f32_l", matmul_id_q4_0_f32_fp32_len, matmul_id_q4_0_f32_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q4_0]->m, "matmul_id_q4_0_f32_m", matmul_id_q4_0_f32_fp32_len, matmul_id_q4_0_f32_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q4_0]->s, "matmul_id_q4_0_f32_s", matmul_id_q4_0_f32_fp32_len, matmul_id_q4_0_f32_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q4_0]->a_l, "matmul_id_q4_0_f32_aligned_l", matmul_id_q4_0_f32_aligned_fp32_len, matmul_id_q4_0_f32_aligned_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q4_0]->a_m, "matmul_id_q4_0_f32_aligned_m", matmul_id_q4_0_f32_aligned_fp32_len, matmul_id_q4_0_f32_aligned_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q4_0]->a_s, "matmul_id_q4_0_f32_aligned_s", matmul_id_q4_0_f32_aligned_fp32_len, matmul_id_q4_0_f32_aligned_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q4_1]->l, "matmul_id_q4_1_f32_l", matmul_id_q4_1_f32_fp32_len, matmul_id_q4_1_f32_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q4_1]->m, "matmul_id_q4_1_f32_m", matmul_id_q4_1_f32_fp32_len, matmul_id_q4_1_f32_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q4_1]->s, "matmul_id_q4_1_f32_s", matmul_id_q4_1_f32_fp32_len, matmul_id_q4_1_f32_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q4_1]->a_l, "matmul_id_q4_1_f32_aligned_l", matmul_id_q4_1_f32_aligned_fp32_len, matmul_id_q4_1_f32_aligned_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q4_1]->a_m, "matmul_id_q4_1_f32_aligned_m", matmul_id_q4_1_f32_aligned_fp32_len, matmul_id_q4_1_f32_aligned_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q4_1]->a_s, "matmul_id_q4_1_f32_aligned_s", matmul_id_q4_1_f32_aligned_fp32_len, matmul_id_q4_1_f32_aligned_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q5_0]->l, "matmul_id_q5_0_f32_l", matmul_id_q5_0_f32_fp32_len, matmul_id_q5_0_f32_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q5_0]->m, "matmul_id_q5_0_f32_m", matmul_id_q5_0_f32_fp32_len, matmul_id_q5_0_f32_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q5_0]->s, "matmul_id_q5_0_f32_s", matmul_id_q5_0_f32_fp32_len, matmul_id_q5_0_f32_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q5_0]->a_l, "matmul_id_q5_0_f32_aligned_l", matmul_id_q5_0_f32_aligned_fp32_len, matmul_id_q5_0_f32_aligned_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q5_0]->a_m, "matmul_id_q5_0_f32_aligned_m", matmul_id_q5_0_f32_aligned_fp32_len, matmul_id_q5_0_f32_aligned_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q5_0]->a_s, "matmul_id_q5_0_f32_aligned_s", matmul_id_q5_0_f32_aligned_fp32_len, matmul_id_q5_0_f32_aligned_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q5_1]->l, "matmul_id_q5_1_f32_l", matmul_id_q5_1_f32_fp32_len, matmul_id_q5_1_f32_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q5_1]->m, "matmul_id_q5_1_f32_m", matmul_id_q5_1_f32_fp32_len, matmul_id_q5_1_f32_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q5_1]->s, "matmul_id_q5_1_f32_s", matmul_id_q5_1_f32_fp32_len, matmul_id_q5_1_f32_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q5_1]->a_l, "matmul_id_q5_1_f32_aligned_l", matmul_id_q5_1_f32_aligned_fp32_len, matmul_id_q5_1_f32_aligned_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q5_1]->a_m, "matmul_id_q5_1_f32_aligned_m", matmul_id_q5_1_f32_aligned_fp32_len, matmul_id_q5_1_f32_aligned_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q5_1]->a_s, "matmul_id_q5_1_f32_aligned_s", matmul_id_q5_1_f32_aligned_fp32_len, matmul_id_q5_1_f32_aligned_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q8_0]->l, "matmul_id_q8_0_f32_l", matmul_id_q8_0_f32_fp32_len, matmul_id_q8_0_f32_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q8_0]->m, "matmul_id_q8_0_f32_m", matmul_id_q8_0_f32_fp32_len, matmul_id_q8_0_f32_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q8_0]->s, "matmul_id_q8_0_f32_s", matmul_id_q8_0_f32_fp32_len, matmul_id_q8_0_f32_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q8_0]->a_l, "matmul_id_q8_0_f32_aligned_l", matmul_id_q8_0_f32_aligned_fp32_len, matmul_id_q8_0_f32_aligned_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q8_0]->a_m, "matmul_id_q8_0_f32_aligned_m", matmul_id_q8_0_f32_aligned_fp32_len, matmul_id_q8_0_f32_aligned_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q8_0]->a_s, "matmul_id_q8_0_f32_aligned_s", matmul_id_q8_0_f32_aligned_fp32_len, matmul_id_q8_0_f32_aligned_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q2_K]->l, "matmul_id_q2_k_f32_l", matmul_id_q2_k_f32_fp32_len, matmul_id_q2_k_f32_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q2_K]->m, "matmul_id_q2_k_f32_m", matmul_id_q2_k_f32_fp32_len, matmul_id_q2_k_f32_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q2_K]->s, "matmul_id_q2_k_f32_s", matmul_id_q2_k_f32_fp32_len, matmul_id_q2_k_f32_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q2_K]->a_l, "matmul_id_q2_k_f32_aligned_l", matmul_id_q2_k_f32_aligned_fp32_len, matmul_id_q2_k_f32_aligned_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q2_K]->a_m, "matmul_id_q2_k_f32_aligned_m", matmul_id_q2_k_f32_aligned_fp32_len, matmul_id_q2_k_f32_aligned_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q2_K]->a_s, "matmul_id_q2_k_f32_aligned_s", matmul_id_q2_k_f32_aligned_fp32_len, matmul_id_q2_k_f32_aligned_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q3_K]->l, "matmul_id_q3_k_f32_l", matmul_id_q3_k_f32_fp32_len, matmul_id_q3_k_f32_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q3_K]->m, "matmul_id_q3_k_f32_m", matmul_id_q3_k_f32_fp32_len, matmul_id_q3_k_f32_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q3_K]->s, "matmul_id_q3_k_f32_s", matmul_id_q3_k_f32_fp32_len, matmul_id_q3_k_f32_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q3_K]->a_l, "matmul_id_q3_k_f32_aligned_l", matmul_id_q3_k_f32_aligned_fp32_len, matmul_id_q3_k_f32_aligned_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q3_K]->a_m, "matmul_id_q3_k_f32_aligned_m", matmul_id_q3_k_f32_aligned_fp32_len, matmul_id_q3_k_f32_aligned_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q3_K]->a_s, "matmul_id_q3_k_f32_aligned_s", matmul_id_q3_k_f32_aligned_fp32_len, matmul_id_q3_k_f32_aligned_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q4_K]->l, "matmul_id_q4_k_f32_l", matmul_id_q4_k_f32_fp32_len, matmul_id_q4_k_f32_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q4_K]->m, "matmul_id_q4_k_f32_m", matmul_id_q4_k_f32_fp32_len, matmul_id_q4_k_f32_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q4_K]->s, "matmul_id_q4_k_f32_s", matmul_id_q4_k_f32_fp32_len, matmul_id_q4_k_f32_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q4_K]->a_l, "matmul_id_q4_k_f32_aligned_l", matmul_id_q4_k_f32_aligned_fp32_len, matmul_id_q4_k_f32_aligned_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q4_K]->a_m, "matmul_id_q4_k_f32_aligned_m", matmul_id_q4_k_f32_aligned_fp32_len, matmul_id_q4_k_f32_aligned_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q4_K]->a_s, "matmul_id_q4_k_f32_aligned_s", matmul_id_q4_k_f32_aligned_fp32_len, matmul_id_q4_k_f32_aligned_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q5_K]->l, "matmul_id_q5_k_f32_l", matmul_id_q5_k_f32_fp32_len, matmul_id_q5_k_f32_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q5_K]->m, "matmul_id_q5_k_f32_m", matmul_id_q5_k_f32_fp32_len, matmul_id_q5_k_f32_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q5_K]->s, "matmul_id_q5_k_f32_s", matmul_id_q5_k_f32_fp32_len, matmul_id_q5_k_f32_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q5_K]->a_l, "matmul_id_q5_k_f32_aligned_l", matmul_id_q5_k_f32_aligned_fp32_len, matmul_id_q5_k_f32_aligned_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q5_K]->a_m, "matmul_id_q5_k_f32_aligned_m", matmul_id_q5_k_f32_aligned_fp32_len, matmul_id_q5_k_f32_aligned_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q5_K]->a_s, "matmul_id_q5_k_f32_aligned_s", matmul_id_q5_k_f32_aligned_fp32_len, matmul_id_q5_k_f32_aligned_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q6_K]->l, "matmul_id_q6_k_f32_l", matmul_id_q6_k_f32_fp32_len, matmul_id_q6_k_f32_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q6_K]->m, "matmul_id_q6_k_f32_m", matmul_id_q6_k_f32_fp32_len, matmul_id_q6_k_f32_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q6_K]->s, "matmul_id_q6_k_f32_s", matmul_id_q6_k_f32_fp32_len, matmul_id_q6_k_f32_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q6_K]->a_l, "matmul_id_q6_k_f32_aligned_l", matmul_id_q6_k_f32_aligned_fp32_len, matmul_id_q6_k_f32_aligned_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), l_wg_denoms, warptile_mmq_l, l_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q6_K]->a_m, "matmul_id_q6_k_f32_aligned_m", matmul_id_q6_k_f32_aligned_fp32_len, matmul_id_q6_k_f32_aligned_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), m_wg_denoms, warptile_mmq_m, m_align);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_mat_id[GGML_TYPE_Q6_K]->a_s, "matmul_id_q6_k_f32_aligned_s", matmul_id_q6_k_f32_aligned_fp32_len, matmul_id_q6_k_f32_aligned_fp32_data, "main", 4, sizeof(vk_mat_mat_id_push_constants), s_wg_denoms, warptile_mmq_s, s_align);
}
// mul mat vec
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_vec_f32_f32[GGML_TYPE_F32 ], "mul_mat_vec_f32_f32_f32", mul_mat_vec_f32_f32_f32_len, mul_mat_vec_f32_f32_f32_data, "main", 3, sizeof(vk_mat_vec_push_constants), {1, 1, 1}, { device->subgroup_size }, 1);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_vec_f32_f32[GGML_TYPE_F16 ], "mul_mat_vec_f16_f32_f32", mul_mat_vec_f16_f32_f32_len, mul_mat_vec_f16_f32_f32_data, "main", 3, sizeof(vk_mat_vec_push_constants), {1, 1, 1}, { device->subgroup_size }, 1);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_vec_f32_f32[GGML_TYPE_Q4_0], "mul_mat_vec_q4_0_f32_f32", mul_mat_vec_q4_0_f32_f32_len, mul_mat_vec_q4_0_f32_f32_data, "main", 3, sizeof(vk_mat_vec_push_constants), {1, 1, 1}, { device->subgroup_size }, 1);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_vec_f32_f32[GGML_TYPE_Q4_1], "mul_mat_vec_q4_1_f32_f32", mul_mat_vec_q4_1_f32_f32_len, mul_mat_vec_q4_1_f32_f32_data, "main", 3, sizeof(vk_mat_vec_push_constants), {1, 1, 1}, { device->subgroup_size }, 1);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_vec_f32_f32[GGML_TYPE_Q5_0], "mul_mat_vec_q5_0_f32_f32", mul_mat_vec_q5_0_f32_f32_len, mul_mat_vec_q5_0_f32_f32_data, "main", 3, sizeof(vk_mat_vec_push_constants), {1, 1, 1}, { device->subgroup_size }, 1);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_vec_f32_f32[GGML_TYPE_Q5_1], "mul_mat_vec_q5_1_f32_f32", mul_mat_vec_q5_1_f32_f32_len, mul_mat_vec_q5_1_f32_f32_data, "main", 3, sizeof(vk_mat_vec_push_constants), {1, 1, 1}, { device->subgroup_size }, 1);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_vec_f32_f32[GGML_TYPE_Q8_0], "mul_mat_vec_q8_0_f32_f32", mul_mat_vec_q8_0_f32_f32_len, mul_mat_vec_q8_0_f32_f32_data, "main", 3, sizeof(vk_mat_vec_push_constants), {1, 1, 1}, { device->subgroup_size }, 1);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_vec_f32_f32[GGML_TYPE_Q2_K], "mul_mat_vec_q2_k_f32_f32", mul_mat_vec_q2_k_f32_f32_len, mul_mat_vec_q2_k_f32_f32_data, "main", 3, sizeof(vk_mat_vec_push_constants), {1, 1, 1}, { device->subgroup_size }, 1);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_vec_f32_f32[GGML_TYPE_Q3_K], "mul_mat_vec_q3_k_f32_f32", mul_mat_vec_q3_k_f32_f32_len, mul_mat_vec_q3_k_f32_f32_data, "main", 3, sizeof(vk_mat_vec_push_constants), {1, 1, 1}, { device->subgroup_size }, 1);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_vec_f32_f32[GGML_TYPE_Q4_K], "mul_mat_vec_q4_k_f32_f32", mul_mat_vec_q4_k_f32_f32_len, mul_mat_vec_q4_k_f32_f32_data, "main", 3, sizeof(vk_mat_vec_push_constants), {1, 1, 1}, { device->subgroup_size }, 1);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_vec_f32_f32[GGML_TYPE_Q5_K], "mul_mat_vec_q5_k_f32_f32", mul_mat_vec_q5_k_f32_f32_len, mul_mat_vec_q5_k_f32_f32_data, "main", 3, sizeof(vk_mat_vec_push_constants), {1, 1, 1}, { device->subgroup_size }, 1);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_vec_f32_f32[GGML_TYPE_Q6_K], "mul_mat_vec_q6_k_f32_f32", mul_mat_vec_q6_k_f32_f32_len, mul_mat_vec_q6_k_f32_f32_data, "main", 3, sizeof(vk_mat_vec_push_constants), {1, 1, 1}, { device->subgroup_size }, 1);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_vec_f16_f32[GGML_TYPE_F32 ], "mul_mat_vec_f32_f16_f32", mul_mat_vec_f32_f16_f32_len, mul_mat_vec_f32_f16_f32_data, "main", 3, sizeof(vk_mat_vec_push_constants), {1, 1, 1}, { device->subgroup_size }, 1);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_vec_f16_f32[GGML_TYPE_F16 ], "mul_mat_vec_f16_f16_f32", mul_mat_vec_f16_f16_f32_len, mul_mat_vec_f16_f16_f32_data, "main", 3, sizeof(vk_mat_vec_push_constants), {1, 1, 1}, { device->subgroup_size }, 1);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_vec_f16_f32[GGML_TYPE_Q4_0], "mul_mat_vec_q4_0_f16_f32", mul_mat_vec_q4_0_f16_f32_len, mul_mat_vec_q4_0_f16_f32_data, "main", 3, sizeof(vk_mat_vec_push_constants), {1, 1, 1}, { device->subgroup_size }, 1);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_vec_f16_f32[GGML_TYPE_Q4_1], "mul_mat_vec_q4_1_f16_f32", mul_mat_vec_q4_1_f16_f32_len, mul_mat_vec_q4_1_f16_f32_data, "main", 3, sizeof(vk_mat_vec_push_constants), {1, 1, 1}, { device->subgroup_size }, 1);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_vec_f16_f32[GGML_TYPE_Q5_0], "mul_mat_vec_q5_0_f16_f32", mul_mat_vec_q5_0_f16_f32_len, mul_mat_vec_q5_0_f16_f32_data, "main", 3, sizeof(vk_mat_vec_push_constants), {1, 1, 1}, { device->subgroup_size }, 1);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_vec_f16_f32[GGML_TYPE_Q5_1], "mul_mat_vec_q5_1_f16_f32", mul_mat_vec_q5_1_f16_f32_len, mul_mat_vec_q5_1_f16_f32_data, "main", 3, sizeof(vk_mat_vec_push_constants), {1, 1, 1}, { device->subgroup_size }, 1);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_vec_f16_f32[GGML_TYPE_Q8_0], "mul_mat_vec_q8_0_f16_f32", mul_mat_vec_q8_0_f16_f32_len, mul_mat_vec_q8_0_f16_f32_data, "main", 3, sizeof(vk_mat_vec_push_constants), {1, 1, 1}, { device->subgroup_size }, 1);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_vec_f16_f32[GGML_TYPE_Q2_K], "mul_mat_vec_q2_k_f16_f32", mul_mat_vec_q2_k_f16_f32_len, mul_mat_vec_q2_k_f16_f32_data, "main", 3, sizeof(vk_mat_vec_push_constants), {1, 1, 1}, { device->subgroup_size }, 1);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_vec_f16_f32[GGML_TYPE_Q3_K], "mul_mat_vec_q3_k_f16_f32", mul_mat_vec_q3_k_f16_f32_len, mul_mat_vec_q3_k_f16_f32_data, "main", 3, sizeof(vk_mat_vec_push_constants), {1, 1, 1}, { device->subgroup_size }, 1);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_vec_f16_f32[GGML_TYPE_Q4_K], "mul_mat_vec_q4_k_f16_f32", mul_mat_vec_q4_k_f16_f32_len, mul_mat_vec_q4_k_f16_f32_data, "main", 3, sizeof(vk_mat_vec_push_constants), {1, 1, 1}, { device->subgroup_size }, 1);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_vec_f16_f32[GGML_TYPE_Q5_K], "mul_mat_vec_q5_k_f16_f32", mul_mat_vec_q5_k_f16_f32_len, mul_mat_vec_q5_k_f16_f32_data, "main", 3, sizeof(vk_mat_vec_push_constants), {1, 1, 1}, { device->subgroup_size }, 1);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_vec_f16_f32[GGML_TYPE_Q6_K], "mul_mat_vec_q6_k_f16_f32", mul_mat_vec_q6_k_f16_f32_len, mul_mat_vec_q6_k_f16_f32_data, "main", 3, sizeof(vk_mat_vec_push_constants), {1, 1, 1}, { device->subgroup_size }, 1);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_vec_id_f32[GGML_TYPE_F32 ], "mul_mat_vec_id_f32_f32", mul_mat_vec_id_f32_f32_len, mul_mat_vec_id_f32_f32_data, "main", 4, sizeof(vk_mat_vec_id_push_constants), {1, 1, 1}, { device->subgroup_size }, 1);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_vec_id_f32[GGML_TYPE_F16 ], "mul_mat_vec_id_f16_f32", mul_mat_vec_id_f16_f32_len, mul_mat_vec_id_f16_f32_data, "main", 4, sizeof(vk_mat_vec_id_push_constants), {1, 1, 1}, { device->subgroup_size }, 1);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_vec_id_f32[GGML_TYPE_Q4_0], "mul_mat_vec_id_q4_0_f32", mul_mat_vec_id_q4_0_f32_len, mul_mat_vec_id_q4_0_f32_data, "main", 4, sizeof(vk_mat_vec_id_push_constants), {1, 1, 1}, { device->subgroup_size }, 1);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_vec_id_f32[GGML_TYPE_Q4_1], "mul_mat_vec_id_q4_1_f32", mul_mat_vec_id_q4_1_f32_len, mul_mat_vec_id_q4_1_f32_data, "main", 4, sizeof(vk_mat_vec_id_push_constants), {1, 1, 1}, { device->subgroup_size }, 1);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_vec_id_f32[GGML_TYPE_Q5_0], "mul_mat_vec_id_q5_0_f32", mul_mat_vec_id_q5_0_f32_len, mul_mat_vec_id_q5_0_f32_data, "main", 4, sizeof(vk_mat_vec_id_push_constants), {1, 1, 1}, { device->subgroup_size }, 1);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_vec_id_f32[GGML_TYPE_Q5_1], "mul_mat_vec_id_q5_1_f32", mul_mat_vec_id_q5_1_f32_len, mul_mat_vec_id_q5_1_f32_data, "main", 4, sizeof(vk_mat_vec_id_push_constants), {1, 1, 1}, { device->subgroup_size }, 1);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_vec_id_f32[GGML_TYPE_Q8_0], "mul_mat_vec_id_q8_0_f32", mul_mat_vec_id_q8_0_f32_len, mul_mat_vec_id_q8_0_f32_data, "main", 4, sizeof(vk_mat_vec_id_push_constants), {1, 1, 1}, { device->subgroup_size }, 1);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_vec_id_f32[GGML_TYPE_Q2_K], "mul_mat_vec_id_q2_k_f32", mul_mat_vec_id_q2_k_f32_len, mul_mat_vec_id_q2_k_f32_data, "main", 4, sizeof(vk_mat_vec_id_push_constants), {1, 1, 1}, { device->subgroup_size }, 1);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_vec_id_f32[GGML_TYPE_Q3_K], "mul_mat_vec_id_q3_k_f32", mul_mat_vec_id_q3_k_f32_len, mul_mat_vec_id_q3_k_f32_data, "main", 4, sizeof(vk_mat_vec_id_push_constants), {1, 1, 1}, { device->subgroup_size }, 1);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_vec_id_f32[GGML_TYPE_Q4_K], "mul_mat_vec_id_q4_k_f32", mul_mat_vec_id_q4_k_f32_len, mul_mat_vec_id_q4_k_f32_data, "main", 4, sizeof(vk_mat_vec_id_push_constants), {1, 1, 1}, { device->subgroup_size }, 1);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_vec_id_f32[GGML_TYPE_Q5_K], "mul_mat_vec_id_q5_k_f32", mul_mat_vec_id_q5_k_f32_len, mul_mat_vec_id_q5_k_f32_data, "main", 4, sizeof(vk_mat_vec_id_push_constants), {1, 1, 1}, { device->subgroup_size }, 1);
ggml_vk_create_pipeline(device, device->pipeline_dequant_mul_mat_vec_id_f32[GGML_TYPE_Q6_K], "mul_mat_vec_id_q6_k_f32", mul_mat_vec_id_q6_k_f32_len, mul_mat_vec_id_q6_k_f32_data, "main", 4, sizeof(vk_mat_vec_id_push_constants), {1, 1, 1}, { device->subgroup_size }, 1);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
// dequant shaders
ggml_vk_create_pipeline(device, device->pipeline_dequant[GGML_TYPE_F32 ], "f32_to_f16", dequant_f32_len, dequant_f32_data, "main", 2, 5 * sizeof(uint32_t), {256 * 16, 1, 1}, {}, 1);
ggml_vk_create_pipeline(device, device->pipeline_dequant[GGML_TYPE_Q4_0], "dequant_q4_0", dequant_q4_0_len, dequant_q4_0_data, "main", 2, 5 * sizeof(uint32_t), {256 * 16, 1, 1}, {}, 1);
ggml_vk_create_pipeline(device, device->pipeline_dequant[GGML_TYPE_Q4_1], "dequant_q4_1", dequant_q4_1_len, dequant_q4_1_data, "main", 2, 5 * sizeof(uint32_t), {256 * 16, 1, 1}, {}, 1);
ggml_vk_create_pipeline(device, device->pipeline_dequant[GGML_TYPE_Q5_0], "dequant_q5_0", dequant_q5_0_len, dequant_q5_0_data, "main", 2, 5 * sizeof(uint32_t), {256 * 16, 1, 1}, {}, 1);
ggml_vk_create_pipeline(device, device->pipeline_dequant[GGML_TYPE_Q5_1], "dequant_q5_1", dequant_q5_1_len, dequant_q5_1_data, "main", 2, 5 * sizeof(uint32_t), {256 * 16, 1, 1}, {}, 1);
ggml_vk_create_pipeline(device, device->pipeline_dequant[GGML_TYPE_Q8_0], "dequant_q8_0", dequant_q8_0_len, dequant_q8_0_data, "main", 2, 5 * sizeof(uint32_t), {256 * 16, 1, 1}, {}, 1);
ggml_vk_create_pipeline(device, device->pipeline_dequant[GGML_TYPE_Q2_K], "dequant_q2_k", dequant_q2_k_len, dequant_q2_k_data, "main", 2, 5 * sizeof(uint32_t), {256 * 64, 1, 1}, {}, 1);
ggml_vk_create_pipeline(device, device->pipeline_dequant[GGML_TYPE_Q3_K], "dequant_q3_k", dequant_q3_k_len, dequant_q3_k_data, "main", 2, 5 * sizeof(uint32_t), {256 * 64, 1, 1}, {}, 1);
ggml_vk_create_pipeline(device, device->pipeline_dequant[GGML_TYPE_Q4_K], "dequant_q4_k", dequant_q4_k_len, dequant_q4_k_data, "main", 2, 5 * sizeof(uint32_t), {256 * 32, 1, 1}, {}, 1);
ggml_vk_create_pipeline(device, device->pipeline_dequant[GGML_TYPE_Q5_K], "dequant_q5_k", dequant_q5_k_len, dequant_q5_k_data, "main", 2, 5 * sizeof(uint32_t), {256 * 64, 1, 1}, {}, 1);
ggml_vk_create_pipeline(device, device->pipeline_dequant[GGML_TYPE_Q6_K], "dequant_q6_k", dequant_q6_k_len, dequant_q6_k_data, "main", 2, 5 * sizeof(uint32_t), {256 * 64, 1, 1}, {}, 1);
// get_rows
ggml_vk_create_pipeline(device, device->pipeline_get_rows[GGML_TYPE_F32 ], "get_rows_f32", get_rows_f32_len, get_rows_f32_data, "main", 3, sizeof(vk_op_binary_push_constants), { 512, 1, 1}, {}, 1);
ggml_vk_create_pipeline(device, device->pipeline_get_rows[GGML_TYPE_F16 ], "get_rows_f16", get_rows_f16_len, get_rows_f16_data, "main", 3, sizeof(vk_op_binary_push_constants), { 512, 1, 1}, {}, 1);
ggml_vk_create_pipeline(device, device->pipeline_get_rows[GGML_TYPE_Q4_0], "get_rows_q4_0", get_rows_q4_0_len, get_rows_q4_0_data, "main", 3, sizeof(vk_op_binary_push_constants), {1024, 1, 1}, {}, 1);
ggml_vk_create_pipeline(device, device->pipeline_get_rows[GGML_TYPE_Q4_1], "get_rows_q4_1", get_rows_q4_1_len, get_rows_q4_1_data, "main", 3, sizeof(vk_op_binary_push_constants), {1024, 1, 1}, {}, 1);
ggml_vk_create_pipeline(device, device->pipeline_get_rows[GGML_TYPE_Q5_0], "get_rows_q5_0", get_rows_q5_0_len, get_rows_q5_0_data, "main", 3, sizeof(vk_op_binary_push_constants), {1024, 1, 1}, {}, 1);
ggml_vk_create_pipeline(device, device->pipeline_get_rows[GGML_TYPE_Q5_1], "get_rows_q5_1", get_rows_q5_1_len, get_rows_q5_1_data, "main", 3, sizeof(vk_op_binary_push_constants), {1024, 1, 1}, {}, 1);
ggml_vk_create_pipeline(device, device->pipeline_get_rows[GGML_TYPE_Q8_0], "get_rows_q8_0", get_rows_q8_0_len, get_rows_q8_0_data, "main", 3, sizeof(vk_op_binary_push_constants), {1024, 1, 1}, {}, 1);
ggml_vk_create_pipeline(device, device->pipeline_get_rows_f32[GGML_TYPE_F32 ], "get_rows_f32_f32", get_rows_f32_f32_len, get_rows_f32_f32_data, "main", 3, sizeof(vk_op_binary_push_constants), { 512, 1, 1}, {}, 1);
ggml_vk_create_pipeline(device, device->pipeline_get_rows_f32[GGML_TYPE_F16 ], "get_rows_f16_f32", get_rows_f16_f32_len, get_rows_f16_f32_data, "main", 3, sizeof(vk_op_binary_push_constants), { 512, 1, 1}, {}, 1);
ggml_vk_create_pipeline(device, device->pipeline_get_rows_f32[GGML_TYPE_Q4_0], "get_rows_q4_0_f32", get_rows_q4_0_f32_len, get_rows_q4_0_f32_data, "main", 3, sizeof(vk_op_binary_push_constants), {1024, 1, 1}, {}, 1);
ggml_vk_create_pipeline(device, device->pipeline_get_rows_f32[GGML_TYPE_Q4_1], "get_rows_q4_1_f32", get_rows_q4_1_f32_len, get_rows_q4_1_f32_data, "main", 3, sizeof(vk_op_binary_push_constants), {1024, 1, 1}, {}, 1);
ggml_vk_create_pipeline(device, device->pipeline_get_rows_f32[GGML_TYPE_Q5_0], "get_rows_q5_0_f32", get_rows_q5_0_f32_len, get_rows_q5_0_f32_data, "main", 3, sizeof(vk_op_binary_push_constants), {1024, 1, 1}, {}, 1);
ggml_vk_create_pipeline(device, device->pipeline_get_rows_f32[GGML_TYPE_Q5_1], "get_rows_q5_1_f32", get_rows_q5_1_f32_len, get_rows_q5_1_f32_data, "main", 3, sizeof(vk_op_binary_push_constants), {1024, 1, 1}, {}, 1);
ggml_vk_create_pipeline(device, device->pipeline_get_rows_f32[GGML_TYPE_Q8_0], "get_rows_q8_0_f32", get_rows_q8_0_f32_len, get_rows_q8_0_f32_data, "main", 3, sizeof(vk_op_binary_push_constants), {1024, 1, 1}, {}, 1);
ggml_vk_create_pipeline(device, device->pipeline_matmul_split_k_reduce, "split_k_reduce", split_k_reduce_len, split_k_reduce_data, "main", 2, 2 * sizeof(uint32_t), {256, 1, 1}, {}, 1);
ggml_vk_create_pipeline(device, device->pipeline_mul_mat_vec_p021_f16_f32, "mul_mat_vec_p021_f16_f32", mul_mat_vec_p021_f16_f32_len, mul_mat_vec_p021_f16_f32_data, "main", 3, 6 * sizeof(uint32_t), {1, 1, 1}, {}, 1);
ggml_vk_create_pipeline(device, device->pipeline_mul_mat_vec_nc_f16_f32, "mul_mat_vec_nc_f16_f32", mul_mat_vec_nc_f16_f32_len, mul_mat_vec_nc_f16_f32_data, "main", 3, 7 * sizeof(uint32_t), {1, 1, 1}, {}, 1);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
ggml_vk_create_pipeline(device, device->pipeline_norm_f32, "norm_f32", norm_f32_len, norm_f32_data, "main", 2, sizeof(vk_op_push_constants), {1, 1, 1}, {}, 1);
ggml_vk_create_pipeline(device, device->pipeline_rms_norm_f32, "rms_norm_f32", rms_norm_f32_len, rms_norm_f32_data, "main", 2, sizeof(vk_op_push_constants), {1, 1, 1}, {}, 1);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
ggml_vk_create_pipeline(device, device->pipeline_cpy_f32_f32, "cpy_f32_f32", cpy_f32_f32_len, cpy_f32_f32_data, "main", 2, sizeof(vk_op_unary_push_constants), {512, 1, 1}, {}, 1);
ggml_vk_create_pipeline(device, device->pipeline_cpy_f32_f16, "cpy_f32_f16", cpy_f32_f16_len, cpy_f32_f16_data, "main", 2, sizeof(vk_op_unary_push_constants), {512, 1, 1}, {}, 1);
ggml_vk_create_pipeline(device, device->pipeline_cpy_f16_f16, "cpy_f16_f16", cpy_f16_f16_len, cpy_f16_f16_data, "main", 2, sizeof(vk_op_unary_push_constants), {512, 1, 1}, {}, 1);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
ggml_vk_create_pipeline(device, device->pipeline_add_f32, "add_f32", add_f32_len, add_f32_data, "main", 3, sizeof(vk_op_binary_push_constants), {512, 1, 1}, {}, 1);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
ggml_vk_create_pipeline(device, device->pipeline_mul_f32, "mul_f32", mul_f32_len, mul_f32_data, "main", 3, sizeof(vk_op_binary_push_constants), {512, 1, 1}, {}, 1);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
ggml_vk_create_pipeline(device, device->pipeline_div_f32, "div_f32", div_f32_len, div_f32_data, "main", 3, sizeof(vk_op_binary_push_constants), {512, 1, 1}, {}, 1);
ggml_vk_create_pipeline(device, device->pipeline_scale_f32, "scale_f32", scale_f32_len, scale_f32_data, "main", 2, sizeof(vk_op_unary_push_constants), {512, 1, 1}, {}, 1);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
ggml_vk_create_pipeline(device, device->pipeline_sqr_f32, "sqr_f32", sqr_f32_len, sqr_f32_data, "main", 2, sizeof(vk_op_unary_push_constants), {512, 1, 1}, {}, 1);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
ggml_vk_create_pipeline(device, device->pipeline_clamp_f32, "clamp_f32", clamp_f32_len, clamp_f32_data, "main", 2, sizeof(vk_op_unary_push_constants), {512, 1, 1}, {}, 1);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
ggml_vk_create_pipeline(device, device->pipeline_gelu_f32, "gelu_f32", gelu_f32_len, gelu_f32_data, "main", 2, sizeof(vk_op_push_constants), {512, 1, 1}, {}, 1);
ggml_vk_create_pipeline(device, device->pipeline_silu_f32, "silu_f32", silu_f32_len, silu_f32_data, "main", 2, sizeof(vk_op_push_constants), {512, 1, 1}, {}, 1);
ggml_vk_create_pipeline(device, device->pipeline_relu_f32, "relu_f32", relu_f32_len, relu_f32_data, "main", 2, sizeof(vk_op_push_constants), {512, 1, 1}, {}, 1);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
ggml_vk_create_pipeline(device, device->pipeline_diag_mask_inf_f32, "diag_mask_inf_f32", diag_mask_inf_f32_len, diag_mask_inf_f32_data, "main", 2, sizeof(vk_op_diag_mask_push_constants), {512, 1, 1}, {}, 1);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
ggml_vk_create_pipeline(device, device->pipeline_soft_max_f32, "soft_max_f32", soft_max_f32_len, soft_max_f32_data, "main", 3, sizeof(vk_op_soft_max_push_constants), {1, 1, 1}, {}, 1);
ggml_vk_create_pipeline(device, device->pipeline_soft_max_f32_f16, "soft_max_f32_f16", soft_max_f32_f16_len, soft_max_f32_f16_data, "main", 3, sizeof(vk_op_soft_max_push_constants), {1, 1, 1}, {}, 1);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
ggml_vk_create_pipeline(device, device->pipeline_rope_norm_f32, "rope_norm_f32", rope_norm_f32_len, rope_norm_f32_data, "main", 4, sizeof(vk_op_rope_push_constants), {1, 512, 1}, {}, 1);
ggml_vk_create_pipeline(device, device->pipeline_rope_norm_f16, "rope_norm_f16", rope_norm_f16_len, rope_norm_f16_data, "main", 4, sizeof(vk_op_rope_push_constants), {1, 512, 1}, {}, 1);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
ggml_vk_create_pipeline(device, device->pipeline_rope_neox_f32, "rope_neox_f32", rope_neox_f32_len, rope_neox_f32_data, "main", 4, sizeof(vk_op_rope_push_constants), {1, 512, 1}, {}, 1);
ggml_vk_create_pipeline(device, device->pipeline_rope_neox_f16, "rope_neox_f16", rope_neox_f16_len, rope_neox_f16_data, "main", 4, sizeof(vk_op_rope_push_constants), {1, 512, 1}, {}, 1);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
ggml_vk_create_pipeline(device, device->pipeline_argsort_f32, "argsort_f32", argsort_f32_len, argsort_f32_data, "main", 2, sizeof(vk_op_argsort_push_constants), {1024, 1, 1}, {}, 1);
ggml_vk_create_pipeline(device, device->pipeline_sum_rows_f32, "sum_rows_f32", sum_rows_f32_len, sum_rows_f32_data, "main", 2, sizeof(vk_op_push_constants), {1, 1, 1}, { device->subgroup_size }, 1);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
static vk_device ggml_vk_get_device(size_t idx) {
VK_LOG_DEBUG("ggml_vk_get_device(" << idx << ")");
if (vk_instance.devices[idx] == nullptr) {
VK_LOG_DEBUG("Initializing new vk_device");
vk_device device = std::make_shared<vk_device_struct>();
vk_instance.devices[idx] = device;
#ifdef GGML_VULKAN_MEMORY_DEBUG
device->memory_logger = std::unique_ptr<vk_memory_logger>(new vk_memory_logger());
#endif
size_t dev_num = vk_instance.device_indices[idx];
std::vector<vk::PhysicalDevice> physical_devices = vk_instance.instance.enumeratePhysicalDevices();
if (dev_num >= physical_devices.size()) {
std::cerr << "ggml_vulkan: Device with index " << dev_num << " does not exist." << std::endl;
throw std::runtime_error("Device not found");
}
device->physical_device = physical_devices[dev_num];
const std::vector<vk::ExtensionProperties> ext_props = device->physical_device.enumerateDeviceExtensionProperties();
bool maintenance4_support = false;
// Check if maintenance4 is supported
for (const auto& properties : ext_props) {
if (strcmp("VK_KHR_maintenance4", properties.extensionName) == 0) {
maintenance4_support = true;
}
}
vk::PhysicalDeviceProperties2 props2;
vk::PhysicalDeviceMaintenance3Properties props3;
vk::PhysicalDeviceMaintenance4Properties props4;
vk::PhysicalDeviceSubgroupProperties subgroup_props;
props2.pNext = &props3;
props3.pNext = &subgroup_props;
if (maintenance4_support) {
subgroup_props.pNext = &props4;
}
device->physical_device.getProperties2(&props2);
device->properties = props2.properties;
const char* GGML_VK_FORCE_MAX_ALLOCATION_SIZE = getenv("GGML_VK_FORCE_MAX_ALLOCATION_SIZE");
if (GGML_VK_FORCE_MAX_ALLOCATION_SIZE != nullptr) {
device->max_memory_allocation_size = std::stoi(GGML_VK_FORCE_MAX_ALLOCATION_SIZE);
} else if (maintenance4_support) {
device->max_memory_allocation_size = std::min(props3.maxMemoryAllocationSize, props4.maxBufferSize);
} else {
device->max_memory_allocation_size = props3.maxMemoryAllocationSize;
}
device->vendor_id = device->properties.vendorID;
device->subgroup_size = subgroup_props.subgroupSize;
device->uma = device->properties.deviceType == vk::PhysicalDeviceType::eIntegratedGpu;
bool fp16_storage = false;
bool fp16_compute = false;
for (const auto& properties : ext_props) {
if (strcmp("VK_KHR_16bit_storage", properties.extensionName) == 0) {
fp16_storage = true;
} else if (strcmp("VK_KHR_shader_float16_int8", properties.extensionName) == 0) {
fp16_compute = true;
}
}
const char* GGML_VK_DISABLE_F16 = getenv("GGML_VK_DISABLE_F16");
const bool force_disable_f16 = GGML_VK_DISABLE_F16 != nullptr;
device->fp16 = !force_disable_f16 && fp16_storage && fp16_compute;
std::vector<vk::QueueFamilyProperties> queue_family_props = device->physical_device.getQueueFamilyProperties();
// Try to find a non-graphics compute queue and transfer-focused queues
const uint32_t compute_queue_family_index = ggml_vk_find_queue_family_index(queue_family_props, vk::QueueFlagBits::eCompute, vk::QueueFlagBits::eGraphics, -1, 1);
const uint32_t transfer_queue_family_index = ggml_vk_find_queue_family_index(queue_family_props, vk::QueueFlagBits::eTransfer, vk::QueueFlagBits::eCompute | vk::QueueFlagBits::eGraphics, compute_queue_family_index, 1);
const float priorities[] = { 1.0f, 1.0f };
device->single_queue = compute_queue_family_index == transfer_queue_family_index && queue_family_props[compute_queue_family_index].queueCount == 1;
std::vector<vk::DeviceQueueCreateInfo> device_queue_create_infos;
if (compute_queue_family_index != transfer_queue_family_index) {
device_queue_create_infos.push_back({vk::DeviceQueueCreateFlags(), compute_queue_family_index, 1, priorities});
device_queue_create_infos.push_back({vk::DeviceQueueCreateFlags(), transfer_queue_family_index, 1, priorities + 1});
} else if(!device->single_queue) {
device_queue_create_infos.push_back({vk::DeviceQueueCreateFlags(), compute_queue_family_index, 2, priorities});
} else {
device_queue_create_infos.push_back({vk::DeviceQueueCreateFlags(), compute_queue_family_index, 1, priorities});
}
vk::DeviceCreateInfo device_create_info;
std::vector<const char *> device_extensions;
vk::PhysicalDeviceFeatures device_features = device->physical_device.getFeatures();
VkPhysicalDeviceFeatures2 device_features2;
device_features2.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_FEATURES_2;
device_features2.pNext = nullptr;
device_features2.features = (VkPhysicalDeviceFeatures)device_features;
VkPhysicalDeviceVulkan11Features vk11_features;
vk11_features.pNext = nullptr;
vk11_features.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_VULKAN_1_1_FEATURES;
device_features2.pNext = &vk11_features;
VkPhysicalDeviceVulkan12Features vk12_features;
vk12_features.pNext = nullptr;
vk12_features.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_VULKAN_1_2_FEATURES;
vk11_features.pNext = &vk12_features;
vkGetPhysicalDeviceFeatures2(device->physical_device, &device_features2);
device->fp16 = device->fp16 && vk12_features.shaderFloat16;
if (!vk11_features.storageBuffer16BitAccess) {
std::cerr << "ggml_vulkan: device " << GGML_VK_NAME << idx << " does not support 16-bit storage." << std::endl;
throw std::runtime_error("Unsupported device");
}
device_extensions.push_back("VK_KHR_16bit_storage");
#ifdef GGML_VULKAN_VALIDATE
device_extensions.push_back("VK_KHR_shader_non_semantic_info");
#endif
if (device->fp16) {
device_extensions.push_back("VK_KHR_shader_float16_int8");
}
device->name = device->properties.deviceName.data();
device_create_info = {
vk::DeviceCreateFlags(),
device_queue_create_infos,
{},
device_extensions
};
device_create_info.setPNext(&device_features2);
device->device = device->physical_device.createDevice(device_create_info);
device->descriptor_set_mode = VK_DEVICE_DESCRIPTOR_POOL_MODE_UNKNOWN;
// Queues
ggml_vk_create_queue(device, device->compute_queue, compute_queue_family_index, 0, { vk::PipelineStageFlagBits::eComputeShader | vk::PipelineStageFlagBits::eTransfer });
// Shaders
ggml_vk_load_shaders(device);
if (!device->single_queue) {
const uint32_t transfer_queue_index = compute_queue_family_index == transfer_queue_family_index ? 1 : 0;
ggml_vk_create_queue(device, device->transfer_queue, transfer_queue_family_index, transfer_queue_index, { vk::PipelineStageFlagBits::eTransfer });
} else {
// TODO: Use pointer or reference to avoid copy
device->transfer_queue = device->compute_queue;
}
device->buffer_type = {
/* .iface = */ ggml_backend_vk_buffer_type_interface,
/* .context = */ new ggml_backend_vk_buffer_type_context{ device->name, device },
};
device->fence = device->device.createFence({});
device->idx = idx;
return device;
}
return vk_instance.devices[idx];
}
static void ggml_vk_print_gpu_info(size_t idx) {
GGML_ASSERT(idx < vk_instance.device_indices.size());
size_t dev_num = vk_instance.device_indices[idx];
VK_LOG_DEBUG("ggml_vk_print_gpu_info(" << dev_num << ")");
GGML_ASSERT(vk_instance_initialized);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
std::vector<vk::PhysicalDevice> devices = vk_instance.instance.enumeratePhysicalDevices();
if (dev_num >= devices.size()) {
std::cerr << "ggml_vulkan: Device with index " << dev_num << " does not exist." << std::endl;
throw std::runtime_error("Device not found");
}
vk::PhysicalDevice physical_device = devices[dev_num];
std::vector<vk::ExtensionProperties> ext_props = physical_device.enumerateDeviceExtensionProperties();
vk::PhysicalDeviceProperties2 props2;
vk::PhysicalDeviceMaintenance3Properties props3;
vk::PhysicalDeviceSubgroupProperties subgroup_props;
vk::PhysicalDeviceDriverProperties driver_props;
props2.pNext = &props3;
props3.pNext = &subgroup_props;
subgroup_props.pNext = &driver_props;
physical_device.getProperties2(&props2);
const size_t subgroup_size = subgroup_props.subgroupSize;
const bool uma = props2.properties.deviceType == vk::PhysicalDeviceType::eIntegratedGpu;
bool fp16_storage = false;
bool fp16_compute = false;
for (auto properties : ext_props) {
if (strcmp("VK_KHR_16bit_storage", properties.extensionName) == 0) {
fp16_storage = true;
} else if (strcmp("VK_KHR_shader_float16_int8", properties.extensionName) == 0) {
fp16_compute = true;
}
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
const char* GGML_VK_DISABLE_F16 = getenv("GGML_VK_DISABLE_F16");
bool force_disable_f16 = GGML_VK_DISABLE_F16 != nullptr;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
bool fp16 = !force_disable_f16 && fp16_storage && fp16_compute;
vk::PhysicalDeviceFeatures device_features = physical_device.getFeatures();
VkPhysicalDeviceFeatures2 device_features2;
device_features2.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_FEATURES_2;
device_features2.pNext = nullptr;
device_features2.features = (VkPhysicalDeviceFeatures)device_features;
VkPhysicalDeviceVulkan11Features vk11_features;
vk11_features.pNext = nullptr;
vk11_features.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_VULKAN_1_1_FEATURES;
device_features2.pNext = &vk11_features;
VkPhysicalDeviceVulkan12Features vk12_features;
vk12_features.pNext = nullptr;
vk12_features.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_VULKAN_1_2_FEATURES;
vk11_features.pNext = &vk12_features;
vkGetPhysicalDeviceFeatures2(physical_device, &device_features2);
fp16 = fp16 && vk12_features.shaderFloat16;
std::string device_name = props2.properties.deviceName.data();
std::cerr << GGML_VK_NAME << idx << ": " << device_name << " (" << driver_props.driverName << ") | uma: " << uma << " | fp16: " << fp16 << " | warp size: " << subgroup_size << std::endl;
if (props2.properties.deviceType == vk::PhysicalDeviceType::eCpu) {
std::cerr << "ggml_vulkan: Warning: Device type is CPU. This is probably not the device you want." << std::endl;
}
}
static bool ggml_vk_instance_validation_ext_available(const std::vector<vk::ExtensionProperties>& instance_extensions);
static bool ggml_vk_instance_portability_enumeration_ext_available(const std::vector<vk::ExtensionProperties>& instance_extensions);
void ggml_vk_instance_init() {
if (vk_instance_initialized) {
return;
}
VK_LOG_DEBUG("ggml_vk_instance_init()");
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
vk_instance_initialized = true;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
vk::ApplicationInfo app_info{ "ggml-vulkan", 1, nullptr, 0, VK_API_VERSION };
const std::vector<vk::ExtensionProperties> instance_extensions = vk::enumerateInstanceExtensionProperties();
const bool validation_ext = ggml_vk_instance_validation_ext_available(instance_extensions);
#ifdef __APPLE__
const bool portability_enumeration_ext = ggml_vk_instance_portability_enumeration_ext_available(instance_extensions);
#endif
std::vector<const char*> layers;
if (validation_ext) {
layers.push_back("VK_LAYER_KHRONOS_validation");
}
std::vector<const char*> extensions;
if (validation_ext) {
extensions.push_back("VK_EXT_validation_features");
}
#ifdef __APPLE__
if (portability_enumeration_ext) {
extensions.push_back("VK_KHR_portability_enumeration");
}
#endif
vk::InstanceCreateInfo instance_create_info(vk::InstanceCreateFlags{}, &app_info, layers, extensions);
#ifdef __APPLE__
if (portability_enumeration_ext) {
instance_create_info.flags |= vk::InstanceCreateFlagBits::eEnumeratePortabilityKHR;
}
#endif
std::vector<vk::ValidationFeatureEnableEXT> features_enable;
vk::ValidationFeaturesEXT validation_features;
if (validation_ext) {
features_enable = { vk::ValidationFeatureEnableEXT::eBestPractices };
validation_features = {
features_enable,
{},
};
validation_features.setPNext(nullptr);
instance_create_info.setPNext(&validation_features);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
std::cerr << "ggml_vulkan: Validation layers enabled" << std::endl;
}
vk_instance.instance = vk::createInstance(instance_create_info);
size_t num_available_devices = vk_instance.instance.enumeratePhysicalDevices().size();
// Emulate behavior of CUDA_VISIBLE_DEVICES for Vulkan
char * devices_env = getenv("GGML_VK_VISIBLE_DEVICES");
if (devices_env != nullptr) {
std::string devices(devices_env);
std::replace(devices.begin(), devices.end(), ',', ' ');
std::stringstream ss(devices);
size_t tmp;
while (ss >> tmp) {
if(tmp >= num_available_devices) {
std::cerr << "ggml_vulkan: Invalid device index " << tmp << " in GGML_VK_VISIBLE_DEVICES." << std::endl;
throw std::runtime_error("Invalid Vulkan device index");
}
vk_instance.device_indices.push_back(tmp);
}
} else {
std::vector<vk::PhysicalDevice> devices = vk_instance.instance.enumeratePhysicalDevices();
// Make sure at least one device exists
if (devices.empty()) {
std::cerr << "ggml_vulkan: Error: No devices found." << std::endl;
GGML_ASSERT(false);
}
// Default to using all dedicated GPUs
for (size_t i = 0; i < devices.size(); i++) {
vk::PhysicalDeviceProperties2 new_props;
vk::PhysicalDeviceDriverProperties new_driver;
vk::PhysicalDeviceIDProperties new_id;
new_props.pNext = &new_driver;
new_driver.pNext = &new_id;
devices[i].getProperties2(&new_props);
if (new_props.properties.deviceType == vk::PhysicalDeviceType::eDiscreteGpu) {
// Check if there are two physical devices corresponding to the same GPU
auto old_device = std::find_if(
vk_instance.device_indices.begin(),
vk_instance.device_indices.end(),
[&devices, &new_id](const size_t k){
vk::PhysicalDeviceProperties2 old_props;
vk::PhysicalDeviceIDProperties old_id;
old_props.pNext = &old_id;
devices[k].getProperties2(&old_props);
return std::equal(std::begin(old_id.deviceUUID), std::end(old_id.deviceUUID), std::begin(new_id.deviceUUID));
}
);
if (old_device == vk_instance.device_indices.end()) {
vk_instance.device_indices.push_back(i);
} else {
// There can be two physical devices corresponding to the same GPU if there are 2 different drivers
// This can cause error when splitting layers aross the devices, need to keep only 1
VK_LOG_DEBUG("Device " << i << " and device " << *old_device << " have the same deviceUUID");
vk::PhysicalDeviceProperties2 old_props;
vk::PhysicalDeviceDriverProperties old_driver;
old_props.pNext = &old_driver;
devices[*old_device].getProperties2(&old_props);
std::map<vk::DriverId, int> driver_priorities {};
int old_priority = std::numeric_limits<int>::max();
int new_priority = std::numeric_limits<int>::max();
// Check https://registry.khronos.org/vulkan/specs/1.3-extensions/man/html/VkDriverId.html for the list of driver id
// Smaller number -> higher priority
switch (old_props.properties.vendorID) {
case VK_VENDOR_ID_AMD:
driver_priorities[vk::DriverId::eMesaRadv] = 1;
driver_priorities[vk::DriverId::eAmdOpenSource] = 2;
driver_priorities[vk::DriverId::eAmdProprietary] = 3;
break;
case VK_VENDOR_ID_INTEL:
driver_priorities[vk::DriverId::eIntelOpenSourceMESA] = 1;
driver_priorities[vk::DriverId::eIntelProprietaryWindows] = 2;
break;
case VK_VENDOR_ID_NVIDIA:
driver_priorities[vk::DriverId::eNvidiaProprietary] = 1;
#if defined(VK_API_VERSION_1_3) && VK_HEADER_VERSION >= 235
driver_priorities[vk::DriverId::eMesaNvk] = 2;
#endif
break;
}
if (driver_priorities.count(old_driver.driverID)) {
old_priority = driver_priorities[old_driver.driverID];
}
if (driver_priorities.count(new_driver.driverID)) {
new_priority = driver_priorities[new_driver.driverID];
}
if (new_priority < old_priority) {
auto r = std::remove(vk_instance.device_indices.begin(), vk_instance.device_indices.end(), *old_device);
vk_instance.device_indices.erase(r, vk_instance.device_indices.end());
vk_instance.device_indices.push_back(i);
VK_LOG_DEBUG("Prioritize device " << i << " driver " << new_driver.driverName << " over device " << *old_device << " driver " << old_driver.driverName);
}
else {
VK_LOG_DEBUG("Prioritize device " << *old_device << " driver " << old_driver.driverName << " over device " << i << " driver " << new_driver.driverName << std::endl);
}
}
}
}
// If no dedicated GPUs found, fall back to GPU 0
if (vk_instance.device_indices.empty()) {
vk_instance.device_indices.push_back(0);
}
}
std::cerr << "ggml_vulkan: Found " << vk_instance.device_indices.size() << " Vulkan devices:" << std::endl;
for (size_t i = 0; i < vk_instance.device_indices.size(); i++) {
ggml_vk_print_gpu_info(i);
}
}
static void ggml_vk_init(ggml_backend_vk_context * ctx, size_t idx) {
GGML_ASSERT(idx < vk_instance.device_indices.size());
VK_LOG_DEBUG("ggml_vk_init(" << ctx->name << ", " << idx << ")");
ggml_vk_instance_init();
ctx->name = GGML_VK_NAME + std::to_string(idx);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
ctx->device = ggml_vk_get_device(idx);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
ctx->semaphore_idx = 0;
ctx->event_idx = 0;
ctx->prealloc_size_x = 0;
ctx->prealloc_size_y = 0;
ctx->prealloc_size_split_k = 0;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
ctx->fence = ctx->device->device.createFence({});
ctx->staging_size = 0;
ctx->staging_offset = 0;
ctx->compute_ctx = nullptr;
ctx->transfer_ctx = nullptr;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
#ifdef GGML_VULKAN_CHECK_RESULTS
const char* skip_checks = getenv("GGML_VULKAN_SKIP_CHECKS");
vk_skip_checks = (skip_checks == NULL ? 0 : atoi(skip_checks));
const char* output_tensor = getenv("GGML_VULKAN_OUTPUT_TENSOR");
vk_output_tensor = (output_tensor == NULL ? 0 : atoi(output_tensor));
#endif
}
static vk_pipeline ggml_vk_get_to_fp16(ggml_backend_vk_context * ctx, ggml_type type) {
VK_LOG_DEBUG("ggml_vk_get_to_fp16()");
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
switch (type) {
case GGML_TYPE_F32:
case GGML_TYPE_Q4_0:
case GGML_TYPE_Q4_1:
case GGML_TYPE_Q5_0:
case GGML_TYPE_Q5_1:
case GGML_TYPE_Q8_0:
case GGML_TYPE_Q2_K:
case GGML_TYPE_Q3_K:
case GGML_TYPE_Q4_K:
case GGML_TYPE_Q5_K:
case GGML_TYPE_Q6_K:
break;
default:
return nullptr;
}
return ctx->device->pipeline_dequant[type];
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
static vk_matmul_pipeline ggml_vk_get_mul_mat_mat_pipeline(ggml_backend_vk_context * ctx, ggml_type src0_type, ggml_type src1_type) {
VK_LOG_DEBUG("ggml_vk_get_mul_mat_mat_pipeline()");
if (src0_type == GGML_TYPE_F32 && src1_type == GGML_TYPE_F32) {
return ctx->device->pipeline_matmul_f32;
}
if (src0_type == GGML_TYPE_F32 && src1_type == GGML_TYPE_F16) {
return ctx->device->pipeline_matmul_f32_f16;
}
if (src0_type == GGML_TYPE_F16 && src1_type == GGML_TYPE_F32) {
return ctx->device->pipeline_matmul_f16_f32;
}
if (src0_type == GGML_TYPE_F16 && src1_type == GGML_TYPE_F16) {
return ctx->device->pipeline_matmul_f16;
}
GGML_ASSERT(src1_type == GGML_TYPE_F32);
switch (src0_type) {
case GGML_TYPE_Q4_0:
case GGML_TYPE_Q4_1:
case GGML_TYPE_Q5_0:
case GGML_TYPE_Q5_1:
case GGML_TYPE_Q8_0:
case GGML_TYPE_Q2_K:
case GGML_TYPE_Q3_K:
case GGML_TYPE_Q4_K:
case GGML_TYPE_Q5_K:
case GGML_TYPE_Q6_K:
break;
default:
return nullptr;
}
return ctx->device->pipeline_dequant_mul_mat_mat[src0_type];
}
static vk_pipeline ggml_vk_get_dequantize_mul_mat_vec(ggml_backend_vk_context * ctx, ggml_type a_type, ggml_type b_type) {
VK_LOG_DEBUG("ggml_vk_get_dequantize_mul_mat_vec()");
GGML_ASSERT(b_type == GGML_TYPE_F32 || b_type == GGML_TYPE_F16);
switch (a_type) {
case GGML_TYPE_F32:
case GGML_TYPE_F16:
case GGML_TYPE_Q4_0:
case GGML_TYPE_Q4_1:
case GGML_TYPE_Q5_0:
case GGML_TYPE_Q5_1:
case GGML_TYPE_Q8_0:
case GGML_TYPE_Q2_K:
case GGML_TYPE_Q3_K:
case GGML_TYPE_Q4_K:
case GGML_TYPE_Q5_K:
case GGML_TYPE_Q6_K:
break;
default:
return nullptr;
}
return b_type == GGML_TYPE_F32 ? ctx->device->pipeline_dequant_mul_mat_vec_f32_f32[a_type] : ctx->device->pipeline_dequant_mul_mat_vec_f16_f32[a_type];
}
static vk_matmul_pipeline ggml_vk_get_mul_mat_mat_id_pipeline(ggml_backend_vk_context * ctx, ggml_type src0_type, ggml_type src1_type) {
VK_LOG_DEBUG("ggml_vk_get_mul_mat_mat_id_pipeline()");
if (src0_type == GGML_TYPE_F32 && src1_type == GGML_TYPE_F32) {
return ctx->device->pipeline_matmul_id_f32;
}
if (src0_type == GGML_TYPE_F16 && src1_type == GGML_TYPE_F32) {
return ctx->device->pipeline_matmul_id_f16_f32;
}
if (src0_type == GGML_TYPE_F16 && src1_type == GGML_TYPE_F16) {
return ctx->device->pipeline_matmul_id_f16;
}
GGML_ASSERT(src1_type == GGML_TYPE_F32);
switch (src0_type) {
case GGML_TYPE_Q4_0:
case GGML_TYPE_Q4_1:
case GGML_TYPE_Q5_0:
case GGML_TYPE_Q5_1:
case GGML_TYPE_Q8_0:
case GGML_TYPE_Q2_K:
case GGML_TYPE_Q3_K:
case GGML_TYPE_Q4_K:
case GGML_TYPE_Q5_K:
case GGML_TYPE_Q6_K:
break;
default:
return nullptr;
}
return ctx->device->pipeline_dequant_mul_mat_mat_id[src0_type];
}
static vk_pipeline ggml_vk_get_dequantize_mul_mat_vec_id(ggml_backend_vk_context * ctx, ggml_type a_type, ggml_type b_type) {
VK_LOG_DEBUG("ggml_vk_get_dequantize_mul_mat_vec()");
GGML_ASSERT(b_type == GGML_TYPE_F32);
switch (a_type) {
case GGML_TYPE_F32:
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
case GGML_TYPE_F16:
case GGML_TYPE_Q4_0:
case GGML_TYPE_Q4_1:
case GGML_TYPE_Q5_0:
case GGML_TYPE_Q5_1:
case GGML_TYPE_Q8_0:
case GGML_TYPE_Q2_K:
case GGML_TYPE_Q3_K:
case GGML_TYPE_Q4_K:
case GGML_TYPE_Q5_K:
case GGML_TYPE_Q6_K:
break;
default:
return nullptr;
}
return ctx->device->pipeline_dequant_mul_mat_vec_id_f32[a_type];
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
static vk_buffer ggml_vk_pool_malloc(ggml_backend_vk_context * ctx, size_t size) {
VK_LOG_DEBUG("ggml_vk_pool_malloc(" << size << ")");
VK_LOG_MEMORY("ggml_vk_pool_malloc");
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
int best_i = -1;
size_t best_size = std::numeric_limits<size_t>::max(); //smallest unused buffer that fits our needs
int worst_i = -1;
size_t worst_size = 0; //largest unused buffer seen so far
for (int i = 0; i < MAX_VK_BUFFERS; ++i) {
vk_buffer &b = ctx->buffer_pool[i];
if (b != nullptr && b->size >= size && b->size < best_size) {
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
best_i = i;
best_size = b->size;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
if (b != nullptr && b->size > worst_size) {
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
worst_i = i;
worst_size = b->size;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
}
if(best_i != -1) {
//found the smallest buffer that fits our needs
vk_buffer b = ctx->buffer_pool[best_i];
ctx->buffer_pool[best_i].reset();
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
return b;
}
if(worst_i != -1) {
//no buffer that fits our needs, resize largest one to save memory
vk_buffer& b = ctx->buffer_pool[worst_i];
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
ggml_vk_destroy_buffer(b);
}
return ggml_vk_create_buffer_device(ctx->device, size);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
static void ggml_vk_pool_free(ggml_backend_vk_context * ctx, vk_buffer& buffer) {
VK_LOG_DEBUG("ggml_vk_pool_free(" << buffer->size << ")");
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
for (int i = 0; i < MAX_VK_BUFFERS; ++i) {
vk_buffer& b = ctx->buffer_pool[i];
if (b == nullptr) {
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
b = buffer;
return;
}
}
std::cerr << "ggml_vulkan: WARNING: vk buffer pool full, increase MAX_VK_BUFFERS" << std::endl;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
ggml_vk_destroy_buffer(buffer);
}
// Returns an available temporary buffer that may only be used temporarily, it will be reused
static vk_buffer ggml_vk_create_buffer_temp(ggml_backend_vk_context * ctx, size_t size) {
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
// Try to find existing temp buffer with enough capacity
for (auto& buffer : ctx->gc.temp_buffers) {
if (buffer->size >= size) {
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
return buffer;
}
}
VK_LOG_MEMORY("ggml_vk_create_buffer_temp(" << size << ")");
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
// Otherwise create new buffer
vk_buffer buf = ggml_vk_pool_malloc(ctx, size);
ctx->gc.temp_buffers.push_back(buf);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
return buf;
}
static void * ggml_vk_host_malloc(vk_device& device, size_t size) {
VK_LOG_MEMORY("ggml_vk_host_malloc(" << size << ")");
vk_buffer buf = ggml_vk_create_buffer(device, size,
vk::MemoryPropertyFlagBits::eHostVisible | vk::MemoryPropertyFlagBits::eHostCoherent | vk::MemoryPropertyFlagBits::eHostCached,
vk::MemoryPropertyFlagBits::eHostVisible | vk::MemoryPropertyFlagBits::eHostCoherent);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
if(!(buf->memory_property_flags & vk::MemoryPropertyFlagBits::eHostVisible)) {
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
fprintf(stderr, "WARNING: failed to allocate %.2f MB of pinned memory\n",
size/1024.0/1024.0);
device->device.freeMemory(buf->device_memory);
device->device.destroyBuffer(buf->buffer);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
return nullptr;
}
device->pinned_memory.push_back(std::make_tuple(buf->ptr, size, buf));
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
return buf->ptr;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
static void ggml_vk_host_free(vk_device& device, void* ptr) {
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
if (ptr == nullptr) {
return;
}
VK_LOG_MEMORY("ggml_vk_host_free(" << ptr << ")");
vk_buffer buf;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
size_t index;
for (size_t i = 0; i < device->pinned_memory.size(); i++) {
const uint8_t* addr = (const uint8_t*) std::get<0>(device->pinned_memory[i]);
const uint8_t* endr = addr + std::get<1>(device->pinned_memory[i]);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
if (ptr >= addr && ptr < endr) {
buf = std::get<2>(device->pinned_memory[i]);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
index = i;
break;
}
}
if (buf == nullptr) {
fprintf(stderr, "WARNING: failed to free pinned memory: memory not in map\n");
return;
}
ggml_vk_destroy_buffer(buf);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
device->pinned_memory.erase(device->pinned_memory.begin() + index);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
static void ggml_vk_host_get(vk_device& device, const void * ptr, vk_buffer& buf, size_t& buf_offset) {
buf = nullptr;
buf_offset = 0;
for (size_t i = 0; i < device->pinned_memory.size(); i++) {
const uint8_t* addr = (const uint8_t*) std::get<0>(device->pinned_memory[i]);
const uint8_t* endr = addr + std::get<1>(device->pinned_memory[i]);
if (ptr >= addr && ptr < endr) {
buf = std::get<2>(device->pinned_memory[i]);
buf_offset = ((const uint8_t *)ptr) - addr;
break;
}
}
}
static vk_submission ggml_vk_begin_submission(vk_device& device, vk_queue& q, bool one_time = true) {
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
vk_submission s;
s.buffer = ggml_vk_create_cmd_buffer(device, q);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
if (one_time) {
s.buffer.begin({ vk::CommandBufferUsageFlagBits::eOneTimeSubmit });
} else {
s.buffer.begin({ vk::CommandBufferUsageFlags{} });
}
return s;
}
static void ggml_vk_dispatch_pipeline(ggml_backend_vk_context * ctx, vk_context * subctx, vk_pipeline& pipeline, std::vector<vk_subbuffer>&& buffers, size_t push_constant_size, const void* push_constants, std::array<uint32_t, 3> elements) {
const uint32_t wg0 = CEIL_DIV(elements[0], pipeline->wg_denoms[0]);
const uint32_t wg1 = CEIL_DIV(elements[1], pipeline->wg_denoms[1]);
const uint32_t wg2 = CEIL_DIV(elements[2], pipeline->wg_denoms[2]);
VK_LOG_DEBUG("ggml_vk_dispatch_pipeline(" << pipeline->name << ", {";
for (auto& buffer : buffers) {
std::cerr << "(" << buffer.buffer << ", " << buffer.offset << ", " << buffer.size << "), ";
}
std::cerr << "}, (" << wg0 << "," << wg1 << "," << wg2 << "))");
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
std::vector<vk::DescriptorBufferInfo> descriptor_buffer_infos;
std::vector<vk::WriteDescriptorSet> write_descriptor_sets;
GGML_ASSERT(pipeline->descriptor_set_idx < pipeline->descriptor_sets.size());
GGML_ASSERT(buffers.size() == pipeline->parameter_count);
vk::DescriptorSet& descriptor_set = pipeline->descriptor_sets[pipeline->descriptor_set_idx++];
for (uint32_t i = 0; i < pipeline->parameter_count; i++) {
descriptor_buffer_infos.push_back({buffers[i].buffer->buffer, buffers[i].offset, buffers[i].size});
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
for (uint32_t i = 0; i < pipeline->parameter_count; i++) {
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
write_descriptor_sets.push_back({descriptor_set, i, 0, 1, vk::DescriptorType::eStorageBuffer, nullptr, &descriptor_buffer_infos[i]});
}
ctx->device->device.updateDescriptorSets(write_descriptor_sets, {});
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
subctx->s->buffer.pushConstants(pipeline->layout, vk::ShaderStageFlagBits::eCompute, 0, push_constant_size, push_constants);
subctx->s->buffer.bindPipeline(vk::PipelineBindPoint::eCompute, pipeline->pipeline);
subctx->s->buffer.bindDescriptorSets(vk::PipelineBindPoint::eCompute,
pipeline->layout,
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
0,
{ descriptor_set },
{});
subctx->s->buffer.dispatch(wg0, wg1, wg2);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
static void ggml_vk_end_submission(vk_submission& s, std::vector<vk_semaphore> wait_semaphores, std::vector<vk_semaphore> signal_semaphores) {
s.buffer.end();
s.wait_semaphores = std::move(wait_semaphores);
s.signal_semaphores = std::move(signal_semaphores);
}
static void ggml_vk_ctx_end(vk_context * ctx) {
VK_LOG_DEBUG("ggml_vk_ctx_end(" << ctx << ", " << ctx->seqs.size() << ")");
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
if (ctx->s == nullptr) {
return;
}
ctx->s->buffer.end();
ctx->s = nullptr;
}
static void ggml_vk_ctx_begin(vk_device& device, vk_context * subctx) {
VK_LOG_DEBUG("ggml_vk_ctx_begin(" << device->name << ")");
if (subctx->s != nullptr) {
ggml_vk_ctx_end(subctx);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
subctx->seqs.push_back({ ggml_vk_begin_submission(device, *subctx->q) });
subctx->s = subctx->seqs[subctx->seqs.size() - 1].data();
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
static size_t ggml_vk_align_size(size_t width, size_t align) {
VK_LOG_DEBUG("ggml_vk_align_size(" << width << ", " << align << ")");
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
return CEIL_DIV(width, align) * align;
}
static void deferred_memcpy(void * dst, const void * src, size_t size, std::vector<vk_staging_memcpy>* memcpys = nullptr) {
if (memcpys == nullptr) {
memcpy(dst, src, size);
} else {
memcpys->emplace_back(dst, src, size);
}
}
static void ggml_vk_ensure_sync_staging_buffer(vk_device& device, size_t size) {
if (device->sync_staging == nullptr || device->sync_staging->size < size) {
VK_LOG_MEMORY("ggml_vk_ensure_sync_staging_buffer(" << size << ")");
ggml_vk_destroy_buffer(device->sync_staging);
device->sync_staging = ggml_vk_create_buffer_check(device, size,
vk::MemoryPropertyFlagBits::eHostVisible | vk::MemoryPropertyFlagBits::eHostCoherent | vk::MemoryPropertyFlagBits::eHostCached,
vk::MemoryPropertyFlagBits::eHostVisible | vk::MemoryPropertyFlagBits::eHostCoherent);
}
}
static void ggml_vk_buffer_write_nc_async(ggml_backend_vk_context * ctx, vk_context * subctx, vk_buffer& dst, size_t offset, const ggml_tensor * tensor, bool sync_staging = false) {
VK_LOG_DEBUG("ggml_vk_buffer_write_nc_async(" << tensor << ")");
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
GGML_ASSERT(!ggml_is_contiguous(tensor));
// Buffer is already mapped
if(dst->memory_property_flags & vk::MemoryPropertyFlagBits::eHostVisible) {
std::cerr << "ggml_vulkan: buffer_write_nc_async dst buffer is host_visible. Use synchronous write." << std::endl;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
GGML_ASSERT(false);
}
// Check if src is pinned memory
vk_buffer buf;
size_t buf_offset;
ggml_vk_host_get(ctx->device, tensor->data, buf, buf_offset);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
const uint64_t ne0 = tensor->ne[0];
const uint64_t ne1 = tensor->ne[1];
const uint64_t ne2 = tensor->ne[2];
const uint64_t ne3 = tensor->ne[3];
const uint64_t nb0 = tensor->nb[0];
const uint64_t nb1 = tensor->nb[1];
const uint64_t nb2 = tensor->nb[2];
const uint64_t nb3 = tensor->nb[3];
const ggml_type type = tensor->type;
const uint64_t ts = ggml_type_size(type);
const uint64_t bs = ggml_blck_size(type);
const uint64_t dstnb0 = ts;
const uint64_t dstnb1 = dstnb0*(ne0/bs);
const uint64_t dstnb2 = dstnb1*ne1;
const uint64_t dstnb3 = dstnb2*ne2;
const uint64_t ne = ggml_nelements(tensor);
if (buf != nullptr) {
// Memory is pinned, use as staging buffer
std::vector<vk::BufferCopy> slices;
for (uint64_t i3 = 0; i3 < ne3; i3++) {
for (uint64_t i2 = 0; i2 < ne2; i2++) {
// Find longest contiguous slice
if (ne1*nb1 == dstnb2) {
slices.push_back({ buf_offset + i3*nb3 + i2*nb2, offset + i3*dstnb3 + i2*dstnb2, dstnb2 });
} else {
for (uint64_t i1 = 0; i1 < ne1; i1++) {
if (ne0*nb0/bs == dstnb1) {
slices.push_back({ buf_offset + i3*nb3 + i2*nb2 + i1*nb1, offset + i3*dstnb3 + i2*dstnb2 + i1*dstnb1, dstnb1 });
} else {
const uint64_t s_off = buf_offset + i3*nb3 + i2*nb2 + i1*nb1;
const uint64_t d_off = offset + i3*dstnb3 + i2*dstnb2 + i1*dstnb1;
for (uint64_t i0 = 0; i0 < ne0; i0++) {
slices.push_back({ s_off + i1*nb0, d_off + i0*dstnb0, dstnb0 });
}
}
}
}
}
}
ggml_vk_sync_buffers(subctx);
subctx->s->buffer.copyBuffer(buf->buffer, dst->buffer, slices);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
return;
}
// Staging buffer required
vk_buffer staging = ctx->staging;
size_t staging_offset = ctx->staging_offset;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
const size_t copy_size = ts*ne/bs;
if (ctx->staging->size < ctx->staging_offset + copy_size) {
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
if (sync_staging) {
// Create temporary larger buffer
ggml_vk_ensure_sync_staging_buffer(ctx->device, copy_size);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
staging = ctx->device->sync_staging;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
staging_offset = 0;
} else {
GGML_ASSERT(false);
}
}
VkBufferCopy buf_copy{ staging_offset, offset, copy_size };
ggml_vk_sync_buffers(subctx);
vkCmdCopyBuffer(subctx->s->buffer, staging->buffer, dst->buffer, 1, &buf_copy);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
for (uint64_t i3 = 0; i3 < ne3; i3++) {
for (uint64_t i2 = 0; i2 < ne2; i2++) {
// Find longest contiguous slice
if (ne1*nb1 == dstnb2) {
deferred_memcpy((uint8_t *)staging->ptr + staging_offset + i3*dstnb3 + i2*dstnb2, (const uint8_t *) tensor->data + buf_offset + i3*nb3 + i2*nb2, dstnb2, &subctx->in_memcpys);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
} else {
for (uint64_t i1 = 0; i1 < ne1; i1++) {
if (ne0*nb0/bs == dstnb1) {
deferred_memcpy((uint8_t *)staging->ptr + staging_offset + i3*dstnb3 + i2*dstnb2 + i1*dstnb1, (const uint8_t *) tensor->data + buf_offset + i3*nb3 + i2*nb2 + i1*nb1, dstnb1, &subctx->in_memcpys);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
} else {
const uint64_t s_off = buf_offset + i3*nb3 + i2*nb2 + i1*nb1;
const uint64_t d_off = staging_offset + i3*dstnb3 + i2*dstnb2 + i1*dstnb1;
for (uint64_t i0 = 0; i0 < ne0; i0++) {
deferred_memcpy((uint8_t *)staging->ptr + d_off + i0*dstnb0, (const uint8_t *) tensor->data + s_off + i0*nb0, dstnb0, &subctx->in_memcpys);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
}
}
}
}
}
}
static void ggml_vk_buffer_write_2d_async(vk_context * subctx, vk_buffer& dst, size_t offset, const void * src, size_t spitch, size_t width, size_t height, vk_buffer staging_buffer, size_t staging_offset, bool sync_staging = false) {
VK_LOG_DEBUG("ggml_vk_buffer_write_2d_async(" << width << ", " << height << ")");
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
// Buffer is already mapped
if(dst->memory_property_flags & vk::MemoryPropertyFlagBits::eHostVisible) {
std::cerr << "ggml_vulkan: buffer_write_async dst buffer is host_visible. Use synchronous write." << std::endl;
GGML_ASSERT(false);
}
// Check if src is pinned memory
vk_buffer buf = nullptr;
size_t buf_offset;
ggml_vk_host_get(dst->device, src, buf, buf_offset);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
if (buf != nullptr) {
// Memory is pinned, use as staging buffer
std::vector<vk::BufferCopy> slices(1);
if (width == spitch) {
// Only do single write if stride is equal
slices[0].srcOffset = buf_offset;
slices[0].dstOffset = offset;
slices[0].size = width * height;
} else {
slices.resize(height);
for (size_t i = 0; i < height; i++) {
slices[i].srcOffset = buf_offset + i * spitch;
slices[i].dstOffset = offset + i * width;
slices[i].size = width;
}
}
ggml_vk_sync_buffers(subctx);
subctx->s->buffer.copyBuffer(buf->buffer, dst->buffer, slices);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
return;
}
VK_LOG_DEBUG("STAGING");
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
// Staging buffer required
const size_t copy_size = width*height;
if (staging_buffer == nullptr || staging_buffer->size < staging_offset + copy_size) {
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
if (sync_staging) {
ggml_vk_ensure_sync_staging_buffer(dst->device, copy_size);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
staging_buffer = dst->device->sync_staging;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
staging_offset = 0;
} else {
GGML_ASSERT(false);
}
}
VkBufferCopy buf_copy = {
staging_offset,
offset,
copy_size};
ggml_vk_sync_buffers(subctx);
vkCmdCopyBuffer(subctx->s->buffer, staging_buffer->buffer, dst->buffer, 1, &buf_copy);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
if (width == spitch) {
deferred_memcpy((uint8_t *)staging_buffer->ptr + staging_offset, src, width * height, &subctx->in_memcpys);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
} else {
for (size_t i = 0; i < height; i++) {
deferred_memcpy((uint8_t *)staging_buffer->ptr + staging_offset + i * width, (const uint8_t *) src + i * spitch, width, &subctx->in_memcpys);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
}
}
static void ggml_vk_buffer_write_async(vk_context * subctx, vk_buffer& dst, size_t offset, const void * src, size_t size, vk_buffer staging_buffer, size_t staging_offset, bool sync_staging = false) {
VK_LOG_DEBUG("ggml_vk_buffer_write_async(" << size << ")");
return ggml_vk_buffer_write_2d_async(subctx, dst, offset, src, size, size, 1, staging_buffer, staging_offset, sync_staging);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
static void ggml_vk_buffer_write_2d(vk_buffer& dst, size_t offset, const void * src, size_t spitch, size_t width, size_t height) {
VK_LOG_DEBUG("ggml_vk_buffer_write_2d(" << width << ", " << height << ")");
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
// Buffer is already mapped
if(dst->memory_property_flags & vk::MemoryPropertyFlagBits::eHostVisible) {
GGML_ASSERT(dst->memory_property_flags & vk::MemoryPropertyFlagBits::eHostCoherent);
for (size_t i = 0; i < height; i++) {
memcpy((uint8_t *)dst->ptr + offset + i * width, (const uint8_t *) src + i * spitch, width);
}
} else {
vk_context * subctx = ggml_vk_create_temporary_context(dst->device->transfer_queue);
ggml_vk_ctx_begin(dst->device, subctx);
ggml_vk_buffer_write_2d_async(subctx, dst, offset, src, spitch, width, height, nullptr, 0, true);
ggml_vk_ctx_end(subctx);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
for (auto& cpy : subctx->in_memcpys) {
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
memcpy(cpy.dst, cpy.src, cpy.n);
}
ggml_vk_submit(subctx, dst->device->fence);
VK_CHECK(dst->device->device.waitForFences({ dst->device->fence }, true, UINT64_MAX), "vk_buffer_write_2d waitForFences");
dst->device->device.resetFences({ dst->device->fence });
delete subctx;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
}
static void ggml_vk_buffer_write(vk_buffer& dst, size_t offset, const void * src, size_t size) {
VK_LOG_DEBUG("ggml_vk_buffer_write(" << size << ")");
ggml_vk_buffer_write_2d(dst, offset, src, 0, size, 1);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
static void ggml_vk_buffer_read_2d_async(vk_context * subctx, vk_buffer& src, size_t offset, void * dst, size_t spitch, size_t dpitch, size_t width, size_t height, vk_buffer staging_buffer, size_t staging_offset, bool sync_staging = false) {
VK_LOG_DEBUG("ggml_vk_buffer_read_2d_async(offset=" << offset << ", width=" << width << ", height=" << height << ")");
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
GGML_ASSERT(width > 0);
GGML_ASSERT(height > 0);
GGML_ASSERT(src != nullptr);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
// Check if dst is pinned memory
vk_buffer buf = nullptr;
size_t buf_offset;
ggml_vk_host_get(src->device, dst, buf, buf_offset);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
std::vector<vk::BufferCopy> slices(1);
if (width == spitch && width == dpitch) {
// Only do single write if stride is equal
slices[0].srcOffset = offset;
slices[0].dstOffset = buf_offset;
slices[0].size = width * height;
} else {
slices.resize(height);
for (size_t i = 0; i < height; i++) {
slices[i].srcOffset = offset + i * spitch;
slices[i].dstOffset = buf_offset + i * dpitch;
slices[i].size = width;
}
}
if (buf != nullptr) {
// Memory is pinned, use as staging buffer
ggml_vk_sync_buffers(subctx);
subctx->s->buffer.copyBuffer(src->buffer, buf->buffer, slices);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
return;
}
VK_LOG_DEBUG("STAGING");
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
// Fall back to staging buffer
const size_t copy_size = dpitch * height;
if (staging_buffer == nullptr || staging_buffer->size < staging_offset + copy_size) {
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
if (sync_staging) {
// Create temporary larger buffer
ggml_vk_ensure_sync_staging_buffer(src->device, copy_size);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
staging_buffer = src->device->sync_staging;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
} else {
GGML_ASSERT(false);
}
}
ggml_vk_sync_buffers(subctx);
subctx->s->buffer.copyBuffer(src->buffer, staging_buffer->buffer, slices);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
deferred_memcpy(dst, staging_buffer->ptr, copy_size, &subctx->out_memcpys);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
static void ggml_vk_buffer_read_async(vk_context * subctx, vk_buffer& src, size_t offset, void * dst, size_t size, vk_buffer staging_buffer, size_t staging_offset, bool sync_staging = false) {
return ggml_vk_buffer_read_2d_async(subctx, src, offset, dst, size, size, size, 1, staging_buffer, staging_offset, sync_staging);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
static void ggml_vk_buffer_read(vk_buffer& src, size_t offset, void * dst, size_t size) {
VK_LOG_DEBUG("ggml_vk_buffer_read(" << offset << ", " << size << ")");
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
if(src->memory_property_flags & vk::MemoryPropertyFlagBits::eHostVisible) {
GGML_ASSERT(src->memory_property_flags & vk::MemoryPropertyFlagBits::eHostCoherent);
memcpy(dst, (uint8_t *) src->ptr + offset, size);
} else {
vk_context * subctx = ggml_vk_create_temporary_context(src->device->transfer_queue);
ggml_vk_ctx_begin(src->device, subctx);
ggml_vk_buffer_read_async(subctx, src, offset, dst, size, nullptr, 0, true);
ggml_vk_ctx_end(subctx);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
ggml_vk_submit(subctx, src->device->fence);
VK_CHECK(src->device->device.waitForFences({ src->device->fence }, true, UINT64_MAX), "vk_buffer_read waitForFences");
src->device->device.resetFences({ src->device->fence });
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
for (auto& cpy : subctx->out_memcpys) {
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
memcpy(cpy.dst, cpy.src, cpy.n);
}
delete subctx;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
}
static void ggml_vk_buffer_copy_async(vk_context * ctx, vk_buffer& dst, size_t dst_offset, vk_buffer& src, size_t src_offset, size_t size) {
VK_LOG_DEBUG("ggml_vk_buffer_copy_async(" << size << ")");
// Make sure both buffers are on same device
GGML_ASSERT(src->device == dst->device);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
VkBufferCopy bc{ src_offset, dst_offset, size };
vkCmdCopyBuffer(ctx->s->buffer, src->buffer, dst->buffer, 1, &bc);
}
static void ggml_vk_buffer_copy(vk_buffer& dst, size_t dst_offset, vk_buffer& src, size_t src_offset, size_t size) {
if (src->device == dst->device) {
VK_LOG_DEBUG("ggml_vk_buffer_copy(SINGLE_DEVICE, " << size << ")");
// Copy within the device
vk_context * subctx = ggml_vk_create_temporary_context(src->device->transfer_queue);
ggml_vk_ctx_begin(src->device, subctx);
ggml_vk_buffer_copy_async(subctx, dst, dst_offset, src, src_offset, size);
ggml_vk_ctx_end(subctx);
ggml_vk_submit(subctx, src->device->fence);
VK_CHECK(src->device->device.waitForFences({ src->device->fence }, true, UINT64_MAX), "vk_buffer_copy waitForFences");
src->device->device.resetFences({ src->device->fence });
delete subctx;
} else {
VK_LOG_DEBUG("ggml_vk_buffer_copy(MULTI_DEVICE, " << size << ")");
// Copy device to device
ggml_vk_ensure_sync_staging_buffer(src->device, size);
ggml_vk_ensure_sync_staging_buffer(dst->device, size);
// Copy to src staging buffer
ggml_vk_buffer_copy(src->device->sync_staging, 0, src, src_offset, size);
// memcpy to dst staging buffer
memcpy(dst->device->sync_staging->ptr, src->device->sync_staging->ptr, size);
// Copy to dst buffer
ggml_vk_buffer_copy(dst, dst_offset, dst->device->sync_staging, 0, size);
}
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
static void ggml_vk_buffer_memset(vk_buffer& dst, size_t offset, uint32_t c, size_t size) {
VK_LOG_DEBUG("ggml_vk_buffer_memset(" << offset << ", " << c << ", " << size << ")");
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
vk_context * subctx = ggml_vk_create_temporary_context(dst->device->transfer_queue);
ggml_vk_ctx_begin(dst->device, subctx);
subctx->s->buffer.fillBuffer(dst->buffer, offset, size, c);
ggml_vk_ctx_end(subctx);
ggml_vk_submit(subctx, dst->device->fence);
VK_CHECK(dst->device->device.waitForFences({ dst->device->fence }, true, UINT64_MAX), "vk_memset waitForFences");
dst->device->device.resetFences({ dst->device->fence });
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
delete subctx;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
static uint32_t ggml_vk_guess_split_k(int m, int n, int k) {
VK_LOG_DEBUG("ggml_vk_guess_split_k(" << m << ", " << n << ", " << k << ")");
// if (k > 128 && (m < 128 || n < 128) && m > 2 && n > 2) {
// return 4;
// }
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
return 1;
GGML_UNUSED(m); GGML_UNUSED(n); GGML_UNUSED(k);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
static vk_pipeline ggml_vk_guess_matmul_pipeline_amd(ggml_backend_vk_context * ctx, vk_matmul_pipeline& mmp, int m, int n, bool aligned) {
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
if (m <= 32 || n <= 32) {
return aligned ? mmp->a_s : mmp->s;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
return aligned ? mmp->a_m : mmp->m;
GGML_UNUSED(ctx);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
static vk_pipeline ggml_vk_guess_matmul_pipeline_apple(ggml_backend_vk_context * ctx, vk_matmul_pipeline& mmp, bool aligned) {
return aligned ? mmp->a_m : mmp->m;
GGML_UNUSED(ctx);
}
static vk_pipeline ggml_vk_guess_matmul_pipeline_intel(ggml_backend_vk_context * ctx, vk_matmul_pipeline& mmp, bool aligned) {
return aligned ? mmp->a_s : mmp->s;
GGML_UNUSED(ctx);
}
static vk_pipeline ggml_vk_guess_matmul_pipeline(ggml_backend_vk_context * ctx, vk_matmul_pipeline& mmp, int m, int n, bool aligned) {
VK_LOG_DEBUG("ggml_vk_guess_matmul_pipeline(" << m << ", " << n << ", " << aligned << ")");
switch (ctx->device->vendor_id) {
case VK_VENDOR_ID_AMD:
return ggml_vk_guess_matmul_pipeline_amd(ctx, mmp, m, n, aligned);
case VK_VENDOR_ID_APPLE:
return ggml_vk_guess_matmul_pipeline_apple(ctx, mmp, aligned);
case VK_VENDOR_ID_INTEL:
return ggml_vk_guess_matmul_pipeline_intel(ctx, mmp, aligned);
default:
break;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
if (m <= 32 || n <= 32) {
return aligned ? mmp->a_s : mmp->s;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
if (m <= 64 || n <= 64) {
return aligned ? mmp->a_m : mmp->m;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
return aligned ? mmp->a_l : mmp->l;
}
static uint32_t ggml_vk_guess_matmul_pipeline_align(ggml_backend_vk_context * ctx, vk_matmul_pipeline& mmp, int m, int n) {
VK_LOG_DEBUG("ggml_vk_guess_matmul_pipeline_align(" << m << ", " << n << ")");
return ggml_vk_guess_matmul_pipeline(ctx, mmp, m, n, true)->align;
}
static void ggml_vk_matmul(
ggml_backend_vk_context * ctx, vk_context * subctx, vk_pipeline& pipeline,
vk_subbuffer&& a, vk_subbuffer&& b, vk_subbuffer&& d, vk_subbuffer&& split_k_buffer,
uint32_t m, uint32_t n, uint32_t k, uint32_t stride_a, uint32_t stride_b, uint32_t stride_d,
uint32_t batch_stride_a, uint32_t batch_stride_b, uint32_t batch_stride_d,
uint32_t split_k, uint32_t batch, uint32_t ne02, uint32_t ne12, uint32_t broadcast2, uint32_t broadcast3) {
VK_LOG_DEBUG("ggml_vk_matmul(a: (" << a.buffer->buffer << ", " << a.offset << ", " << a.size << "), b: (" << b.buffer->buffer << ", " << b.offset << ", " << b.size << "), d: (" << d.buffer->buffer << ", " << d.offset << ", " << d.size << "), split_k: (" << (split_k_buffer.buffer != nullptr ? split_k_buffer.buffer->buffer : VK_NULL_HANDLE) << ", " << split_k_buffer.offset << ", " << split_k_buffer.size << "), m: " << m << ", n: " << n << ", k: " << k << ", stride_a: " << stride_a << ", stride_b: " << stride_b << ", stride_d: " << stride_d << ", batch_stride_a: " << batch_stride_a << ", batch_stride_b: " << batch_stride_b << ", batch_stride_d: " << batch_stride_d << ", split_k: " << split_k << ", batch: " << batch << ", ne02: " << ne02 << ", ne12: " << ne12 << ", broadcast2: " << broadcast2 << ", broadcast3: " << broadcast3 << ")");
ggml_vk_sync_buffers(subctx);
if (split_k == 1) {
const vk_mat_mat_push_constants pc = { m, n, k, stride_a, stride_b, stride_d, batch_stride_a, batch_stride_b, batch_stride_d, k, ne02, ne12, broadcast2, broadcast3 };
ggml_vk_dispatch_pipeline(ctx, subctx, pipeline, { a, b, d }, sizeof(vk_mat_mat_push_constants), &pc, { m, n, batch });
return;
}
GGML_ASSERT(batch_stride_d == m * n);
const vk_mat_mat_push_constants pc1 = { m, n, k, stride_a, stride_b, stride_d, batch_stride_a, batch_stride_b, batch_stride_d, CEIL_DIV(k, split_k), ne02, ne12, broadcast2, broadcast3 };
// Make sure enough workgroups get assigned for split k to work
ggml_vk_dispatch_pipeline(ctx, subctx, pipeline, { a, b, split_k_buffer }, sizeof(vk_mat_mat_push_constants), &pc1, { (CEIL_DIV(m, pipeline->wg_denoms[0]) * pipeline->wg_denoms[0]) * split_k, n, batch });
ggml_vk_sync_buffers(subctx);
const std::array<uint32_t, 2> pc2 = { (uint32_t)(m * n * batch), split_k };
ggml_vk_dispatch_pipeline(ctx, subctx, ctx->device->pipeline_matmul_split_k_reduce, { split_k_buffer, d }, pc2.size() * sizeof(uint32_t), pc2.data(), { m * n * batch, 1, 1 });
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
static void ggml_vk_matmul_id(
ggml_backend_vk_context * ctx, vk_context * subctx, vk_pipeline& pipeline,
vk_subbuffer&& a, vk_subbuffer&& b, vk_subbuffer&& d, vk_subbuffer&& ids,
uint32_t m, uint32_t n, uint32_t k, uint32_t stride_a, uint32_t stride_b, uint32_t stride_d,
uint32_t batch_stride_a, uint32_t batch_stride_b, uint32_t batch_stride_d,
uint32_t n_as, uint32_t nei0, uint32_t nei1, uint32_t nbi1, uint32_t ne11) {
VK_LOG_DEBUG("ggml_vk_matmul_id(a: (" << a.buffer->buffer << ", " << a.offset << ", " << a.size << "), b: (" << b.buffer->buffer << ", " << b.offset << ", " << b.size << "), d: (" << d.buffer->buffer << ", " << d.offset << ", " << d.size << "), ids: (" << ids.buffer->buffer << ", " << ids.offset << ", " << ids.size << "), " <<
"m: " << m << ", n: " << n << ", k: " << k << ", stride_a: " << stride_a << ", stride_b: " << stride_b << ", stride_d: " << stride_d << ", " <<
"batch_stride_a: " << batch_stride_a << ", batch_stride_b: " << batch_stride_b << ", batch_stride_d: " << batch_stride_d << ", " <<
"n_as: " << n_as << ", nei0: " << nei0 << ", nei1: " << nei1 << ", nbi1: " << nbi1 << ", ne11: " << ne11 << ")");
ggml_vk_sync_buffers(subctx);
const vk_mat_mat_id_push_constants pc = { m, n, k, stride_a, stride_b, stride_d, batch_stride_a, batch_stride_b, batch_stride_d,
nei0, nei1, nbi1, ne11 };
ggml_vk_dispatch_pipeline(ctx, subctx, pipeline, { a, b, d, ids }, sizeof(vk_mat_mat_id_push_constants), &pc, { m, nei1, n_as });
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
static bool ggml_vk_dim01_contiguous(const ggml_tensor * tensor) {
return
tensor->nb[0] == ggml_type_size(tensor->type) &&
tensor->nb[1] == (tensor->nb[0]*tensor->ne[0])/ggml_blck_size(tensor->type) &&
tensor->nb[3] == tensor->nb[2]*tensor->ne[2];
}
static vk_pipeline ggml_vk_get_cpy_pipeline(ggml_backend_vk_context * ctx, ggml_type from, ggml_type to) {
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
if (from == GGML_TYPE_F32 && to == GGML_TYPE_F32) {
return ctx->device->pipeline_cpy_f32_f32;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
if (from == GGML_TYPE_F32 && to == GGML_TYPE_F16) {
return ctx->device->pipeline_cpy_f32_f16;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
if (from == GGML_TYPE_F16 && to == GGML_TYPE_F16) {
return ctx->device->pipeline_cpy_f16_f16;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
std::cerr << "Missing CPY op for types: " << ggml_type_name(from) << " " << ggml_type_name(to) << std::endl;
GGML_ASSERT(false);
}
static void ggml_vk_cpy_to_contiguous(ggml_backend_vk_context * ctx, vk_context * subctx, vk_pipeline pipeline, const ggml_tensor * tensor, vk_subbuffer&& in, vk_subbuffer&& out) {
VK_LOG_DEBUG("ggml_vk_cpy_to_contiguous((" << tensor << ", type=" << tensor->type << ", ne0=" << tensor->ne[0] << ", ne1=" << tensor->ne[1] << ", ne2=" << tensor->ne[2] << ", ne3=" << tensor->ne[3] << ", nb0=" << tensor->nb[0] << ", nb1=" << tensor->nb[1] << ", nb2=" << tensor->nb[2] << ", nb3=" << tensor->nb[3] << "), ";
std::cerr << "buffer in size=" << in.buffer->size << ", buffer out size=" << out.buffer->size << ")");
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
const int tensor_type_size = ggml_type_size(tensor->type);
const uint32_t ne = ggml_nelements(tensor);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
const vk_op_unary_push_constants pc = {
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
(uint32_t)ne,
(uint32_t)tensor->ne[0], (uint32_t)tensor->ne[1], (uint32_t)tensor->ne[2], (uint32_t)tensor->ne[3], (uint32_t)tensor->nb[0] / tensor_type_size, (uint32_t)tensor->nb[1] / tensor_type_size, (uint32_t)tensor->nb[2] / tensor_type_size, (uint32_t)tensor->nb[3] / tensor_type_size,
(uint32_t)tensor->ne[0], (uint32_t)tensor->ne[1], (uint32_t)tensor->ne[2], (uint32_t)tensor->ne[3], 1 , (uint32_t)tensor->ne[0] , (uint32_t)(tensor->ne[0] * tensor->ne[1]) , (uint32_t)(tensor->ne[0] * tensor->ne[1] * tensor->ne[2]),
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
0,
0.0f, 0.0f,
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
};
ggml_vk_sync_buffers(subctx);
ggml_vk_dispatch_pipeline(ctx, subctx, pipeline, { in, out }, sizeof(vk_op_unary_push_constants), &pc, { ne, 1, 1 });
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
static void ggml_vk_mul_mat_q_f16(ggml_backend_vk_context * ctx, vk_context * subctx, const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst) {
VK_LOG_DEBUG("ggml_vk_mul_mat_q_f16((" << src0 << ", name=" << src0->name << ", type=" << src0->type << ", ne0=" << src0->ne[0] << ", ne1=" << src0->ne[1] << ", ne2=" << src0->ne[2] << ", ne3=" << src0->ne[3] << ", nb0=" << src0->nb[0] << ", nb1=" << src0->nb[1] << ", nb2=" << src0->nb[2] << ", nb3=" << src0->nb[3];
std::cerr << "), (" << src1 << ", name=" << src1->name << ", type=" << src1->type << ", ne0=" << src1->ne[0] << ", ne1=" << src1->ne[1] << ", ne2=" << src1->ne[2] << ", ne3=" << src1->ne[3] << ", nb0=" << src1->nb[0] << ", nb1=" << src1->nb[1] << ", nb2=" << src1->nb[2] << ", nb3=" << src1->nb[3];
std::cerr << "), (" << dst << ", name=" << dst->name << ", type=" << dst->type << ", ne0=" << dst->ne[0] << ", ne1=" << dst->ne[1] << ", ne2=" << dst->ne[2] << ", ne3=" << dst->ne[3] << ", nb0=" << dst->nb[0] << ", nb1=" << dst->nb[1] << ", nb2=" << dst->nb[2] << ", nb3=" << dst->nb[3] << "),)");
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
GGML_ASSERT(ggml_vk_dim01_contiguous(src0) || src0->type == GGML_TYPE_F32 || src0->type == GGML_TYPE_F16); // NOLINT
GGML_ASSERT(ggml_vk_dim01_contiguous(src1) || src1->type == GGML_TYPE_F32 || src1->type == GGML_TYPE_F16); // NOLINT
const uint64_t ne00 = src0->ne[0];
const uint64_t ne01 = src0->ne[1];
const uint64_t ne02 = src0->ne[2];
const uint64_t ne03 = src0->ne[3];
const uint64_t ne10 = src1->ne[0];
const uint64_t ne11 = src1->ne[1];
const uint64_t ne12 = src1->ne[2];
const uint64_t ne13 = src1->ne[3];
const uint64_t ne20 = dst->ne[0];
const uint64_t ne21 = dst->ne[1];
const uint64_t r2 = ne12 / ne02;
const uint64_t r3 = ne13 / ne03;
ggml_tensor_extra_gpu * extra = (ggml_tensor_extra_gpu *) dst->extra;
ggml_tensor_extra_gpu * extra_src0 = (ggml_tensor_extra_gpu *) src0->extra;
ggml_tensor_extra_gpu * extra_src1 = (ggml_tensor_extra_gpu *) src1->extra;
vk_buffer d_Qx;
size_t qx_buf_offset = 0;
vk_buffer d_Qy;
size_t qy_buf_offset = 0;
bool src0_uma = false;
bool src1_uma = false;
if (ctx->device->uma) {
ggml_vk_host_get(ctx->device, src0->data, d_Qx, qx_buf_offset);
ggml_vk_host_get(ctx->device, src1->data, d_Qy, qy_buf_offset);
src0_uma = d_Qx != nullptr;
src1_uma = d_Qy != nullptr;
}
const bool x_non_contig = !ggml_vk_dim01_contiguous(src0);
const bool y_non_contig = !ggml_vk_dim01_contiguous(src1);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
const bool y_f32_kernel = src1->type == GGML_TYPE_F32 && !y_non_contig;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
vk_matmul_pipeline mmp = ggml_vk_get_mul_mat_mat_pipeline(ctx, src0->type, y_non_contig ? GGML_TYPE_F16 : src1->type);
const bool qx_needs_dequant = mmp == nullptr || x_non_contig;
const bool qy_needs_dequant = (src1->type != GGML_TYPE_F16 && !y_f32_kernel) || y_non_contig;
if (mmp == nullptr) {
// Fall back to dequant + f16 mulmat
mmp = ggml_vk_get_mul_mat_mat_pipeline(ctx, GGML_TYPE_F16, y_f32_kernel ? GGML_TYPE_F32 : GGML_TYPE_F16);
}
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
// Not implemented
GGML_ASSERT(y_non_contig || !qy_needs_dequant); // NOLINT
const int x_ne = ne01 * ne00;
const int y_ne = ne11 * ne10;
const int d_ne = ne11 * ne01;
const uint32_t kpad = ggml_vk_align_size(ne10, ggml_vk_guess_matmul_pipeline_align(ctx, mmp, ne01, ne11));
const bool aligned = ne10 == kpad && ne01 > 8 && ne11 > 8;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
const uint32_t split_k = ggml_vk_guess_split_k(ne01, ne11, ne10);
vk_pipeline pipeline = ggml_vk_guess_matmul_pipeline(ctx, mmp, ne01, ne11, aligned);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
const uint64_t qx_sz = ggml_type_size(src0->type) * x_ne / ggml_blck_size(src0->type);
const uint64_t qy_sz = ggml_type_size(src1->type) * y_ne / ggml_blck_size(src1->type);
const uint64_t x_sz = !qx_needs_dequant ? qx_sz : sizeof(ggml_fp16_t) * x_ne;
const uint64_t y_sz = y_f32_kernel ? sizeof(float) * y_ne : sizeof(ggml_fp16_t) * y_ne;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
const uint64_t d_sz = sizeof(float) * d_ne;
vk_buffer d_D = extra->buffer_gpu.lock();
const uint64_t d_buf_offset = extra->offset + dst->view_offs;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
GGML_ASSERT(d_D != nullptr);
GGML_ASSERT(d_D->size >= d_buf_offset + d_sz * ne02 * ne03);
vk_buffer d_X;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
uint64_t x_buf_offset = 0;
vk_buffer d_Y;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
uint64_t y_buf_offset = 0;
if (!src0_uma) {
d_Qx = extra_src0->buffer_gpu.lock();
qx_buf_offset = extra_src0->offset + src0->view_offs;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
GGML_ASSERT(d_Qx != nullptr);
}
if (!src1_uma) {
d_Qy = extra_src1->buffer_gpu.lock();
qy_buf_offset = extra_src1->offset + src1->view_offs;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
GGML_ASSERT(d_Qy != nullptr);
}
if (qx_needs_dequant) {
d_X = ctx->prealloc_x;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
GGML_ASSERT(d_X->size >= x_sz * ne02 * ne03);
} else {
d_X = d_Qx;
x_buf_offset = qx_buf_offset;
GGML_ASSERT(qx_sz == x_sz);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
if (qy_needs_dequant) {
d_Y = ctx->prealloc_y;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
GGML_ASSERT(d_Y->size >= y_sz * ne02 * ne03);
} else {
d_Y = d_Qy;
y_buf_offset = qy_buf_offset;
GGML_ASSERT(qy_sz == y_sz);
}
vk_pipeline to_fp16_vk_0 = nullptr;
vk_pipeline to_fp16_vk_1 = nullptr;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
if (x_non_contig) {
to_fp16_vk_0 = ggml_vk_get_cpy_pipeline(ctx, src0->type, GGML_TYPE_F16);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
} else {
to_fp16_vk_0 = ggml_vk_get_to_fp16(ctx, src0->type);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
if (y_non_contig) {
to_fp16_vk_1 = ggml_vk_get_cpy_pipeline(ctx, src1->type, GGML_TYPE_F16);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
} else {
to_fp16_vk_1 = ggml_vk_get_to_fp16(ctx, src1->type);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
GGML_ASSERT(!qx_needs_dequant || to_fp16_vk_0 != nullptr); // NOLINT
GGML_ASSERT(!qy_needs_dequant || to_fp16_vk_1 != nullptr); // NOLINT
// Allocate descriptor sets
ggml_pipeline_allocate_descriptor_sets(ctx->device, pipeline, 1);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
if (qx_needs_dequant) {
ggml_pipeline_allocate_descriptor_sets(ctx->device, to_fp16_vk_0, 1);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
if (qy_needs_dequant) {
ggml_pipeline_allocate_descriptor_sets(ctx->device, to_fp16_vk_1, 1);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
if (split_k > 1) {
ggml_pipeline_allocate_descriptor_sets(ctx->device, ctx->device->pipeline_matmul_split_k_reduce, 1);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
if (x_non_contig) {
ggml_vk_cpy_to_contiguous(ctx, subctx, to_fp16_vk_0, src0, { d_Qx, qx_buf_offset, VK_WHOLE_SIZE }, { d_X, 0, VK_WHOLE_SIZE });
} else if (qx_needs_dequant) {
const std::vector<uint32_t> pc = { (uint32_t)ne01, (uint32_t)ne10, (uint32_t)ne10, (uint32_t)ne10, (uint32_t)(ggml_nelements(src0)) };
ggml_vk_sync_buffers(subctx);
ggml_vk_dispatch_pipeline(ctx, subctx, to_fp16_vk_0, { { d_Qx, qx_buf_offset, qx_sz * ne02 * ne03 }, { d_X, 0, x_sz * ne02 * ne03 } }, pc.size() * sizeof(uint32_t), pc.data(), { (uint32_t)(x_ne * ne02 * ne03), 1, 1});
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
if (y_non_contig) {
ggml_vk_cpy_to_contiguous(ctx, subctx, to_fp16_vk_1, src1, { d_Qy, qy_buf_offset, VK_WHOLE_SIZE }, { d_Y, 0, VK_WHOLE_SIZE });
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
uint32_t stride_batch_x = ne00*ne01;
uint32_t stride_batch_y = ne10*ne11;
if (!ggml_vk_dim01_contiguous(src0) && !qx_needs_dequant) {
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
stride_batch_x = src0->nb[0] / ggml_type_size(src0->type);
}
if (!ggml_vk_dim01_contiguous(src1) && !qy_needs_dequant) {
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
stride_batch_y = src1->nb[0] / ggml_type_size(src1->type);
}
// compute
ggml_vk_matmul(
ctx, subctx, pipeline,
{ d_X, x_buf_offset, x_sz * ne02 * ne03 }, { d_Y, y_buf_offset, y_sz * ne12 * ne13 },
{ d_D, d_buf_offset, d_sz * ne12 * ne13 }, { ctx->prealloc_split_k, 0, d_sz * ne12 * ne13 * split_k },
ne01, ne11, ne10,
ne10, ne10, ne01, stride_batch_x, stride_batch_y, ne20*ne21,
split_k, ne12*ne13, ne02, ne12, r2, r3
); // NOLINT
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
static void ggml_vk_mul_mat_vec_q_f16(ggml_backend_vk_context * ctx, vk_context * subctx, const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst) {
VK_LOG_DEBUG("ggml_vk_mul_mat_vec_q_f16((" << src0 << ", name=" << src0->name << ", type=" << src0->type << ", ne0=" << src0->ne[0] << ", ne1=" << src0->ne[1] << ", ne2=" << src0->ne[2] << ", ne3=" << src0->ne[3] << ", nb0=" << src0->nb[0] << ", nb1=" << src0->nb[1] << ", nb2=" << src0->nb[2] << ", nb3=" << src0->nb[3];
std::cerr << "), (" << src1 << ", name=" << src1->name << ", type=" << src1->type << ", ne0=" << src1->ne[0] << ", ne1=" << src1->ne[1] << ", ne2=" << src1->ne[2] << ", ne3=" << src1->ne[3] << ", nb0=" << src1->nb[0] << ", nb1=" << src1->nb[1] << ", nb2=" << src1->nb[2] << ", nb3=" << src1->nb[3];
std::cerr << "), (" << dst << ", name=" << dst->name << ", type=" << dst->type << ", ne0=" << dst->ne[0] << ", ne1=" << dst->ne[1] << ", ne2=" << dst->ne[2] << ", ne3=" << dst->ne[3] << ", nb0=" << dst->nb[0] << ", nb1=" << dst->nb[1] << ", nb2=" << dst->nb[2] << ", nb3=" << dst->nb[3] << "),)");
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
GGML_ASSERT(ggml_vk_dim01_contiguous(src0) || src0->type == GGML_TYPE_F32 || src0->type == GGML_TYPE_F16); // NOLINT
GGML_ASSERT(ggml_vk_dim01_contiguous(src1) || src1->type == GGML_TYPE_F32 || src1->type == GGML_TYPE_F16); // NOLINT
const uint64_t ne00 = src0->ne[0];
const uint64_t ne01 = src0->ne[1];
const uint64_t ne02 = src0->ne[2];
const uint64_t ne03 = src0->ne[3];
const uint64_t ne10 = src1->ne[0];
const uint64_t ne11 = src1->ne[1];
const uint64_t ne12 = src1->ne[2];
const uint64_t ne13 = src1->ne[3];
GGML_ASSERT(ne11 == 1);
const uint64_t ne20 = dst->ne[0];
const uint64_t ne21 = dst->ne[1];
const uint64_t ne22 = dst->ne[2];
const uint64_t ne23 = dst->ne[3];
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
const uint64_t r2 = ne12 / ne02;
const uint64_t r3 = ne13 / ne03;
ggml_tensor_extra_gpu * extra = (ggml_tensor_extra_gpu *) dst->extra;
ggml_tensor_extra_gpu * extra_src0 = (ggml_tensor_extra_gpu *) src0->extra;
ggml_tensor_extra_gpu * extra_src1 = (ggml_tensor_extra_gpu *) src1->extra;
vk_buffer d_Qx;
size_t qx_buf_offset = 0;
vk_buffer d_Qy;
size_t qy_buf_offset = 0;
bool src0_uma = false;
bool src1_uma = false;
if (ctx->device->uma) {
ggml_vk_host_get(ctx->device, src0->data, d_Qx, qx_buf_offset);
ggml_vk_host_get(ctx->device, src1->data, d_Qy, qy_buf_offset);
src0_uma = d_Qx != nullptr;
src1_uma = d_Qy != nullptr;
}
const bool x_non_contig = !ggml_vk_dim01_contiguous(src0);
const bool y_non_contig = !ggml_vk_dim01_contiguous(src1);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
const bool f16_f32_kernel = src1->type == GGML_TYPE_F32;
const bool qx_needs_dequant = x_non_contig;
const bool qy_needs_dequant = (src1->type != GGML_TYPE_F16 && !f16_f32_kernel) || y_non_contig;
// Not implemented
GGML_ASSERT(y_non_contig || !qy_needs_dequant); // NOLINT
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
const uint64_t x_ne = ne01 * ne00;
const uint64_t y_ne = ne11 * ne10;
const uint64_t d_ne = ne11 * ne01;
const uint64_t qx_sz = ggml_vk_align_size(ggml_type_size(src0->type) * x_ne / ggml_blck_size(src0->type), ctx->device->properties.limits.minStorageBufferOffsetAlignment);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
const uint64_t qy_sz = ggml_type_size(src1->type) * y_ne / ggml_blck_size(src1->type);
const uint64_t x_sz = x_non_contig ? ggml_vk_align_size(ggml_type_size(src0->type) * x_ne, ctx->device->properties.limits.minStorageBufferOffsetAlignment) : qx_sz;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
const uint64_t y_sz = f16_f32_kernel ? sizeof(float) * y_ne : sizeof(ggml_fp16_t) * y_ne;
const uint64_t d_sz = sizeof(float) * d_ne;
vk_buffer d_D = extra->buffer_gpu.lock();
const uint64_t d_buf_offset = extra->offset + dst->view_offs;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
GGML_ASSERT(d_D != nullptr);
vk_buffer d_X;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
uint64_t x_buf_offset = 0;
vk_buffer d_Y;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
uint64_t y_buf_offset = 0;
if(!src0_uma) {
d_Qx = extra_src0->buffer_gpu.lock();
qx_buf_offset = extra_src0->offset + src0->view_offs;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
GGML_ASSERT(d_Qx != nullptr);
}
if(!src1_uma) {
d_Qy = extra_src1->buffer_gpu.lock();
qy_buf_offset = extra_src1->offset + src1->view_offs;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
GGML_ASSERT(d_Qy != nullptr);
}
if (qx_needs_dequant) {
d_X = ctx->prealloc_x;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
} else {
d_X = d_Qx;
x_buf_offset = qx_buf_offset;
GGML_ASSERT(qx_sz == x_sz);
}
if (qy_needs_dequant) {
d_Y = ctx->prealloc_y;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
} else {
d_Y = d_Qy;
y_buf_offset = qy_buf_offset;
GGML_ASSERT(qy_sz == y_sz);
}
vk_pipeline to_fp16_vk_0 = nullptr;
vk_pipeline to_fp16_vk_1 = nullptr;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
if (x_non_contig) {
to_fp16_vk_0 = ggml_vk_get_cpy_pipeline(ctx, src0->type, src0->type);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
if (y_non_contig) {
to_fp16_vk_1 = ggml_vk_get_cpy_pipeline(ctx, src1->type, src1->type);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
} else {
to_fp16_vk_1 = ggml_vk_get_to_fp16(ctx, src1->type);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
vk_pipeline dmmv = ggml_vk_get_dequantize_mul_mat_vec(ctx, src0->type, src1->type);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
GGML_ASSERT(!qx_needs_dequant || to_fp16_vk_0 != nullptr); // NOLINT
GGML_ASSERT(!qy_needs_dequant || to_fp16_vk_1 != nullptr); // NOLINT
GGML_ASSERT(dmmv != nullptr);
// Allocate descriptor sets
if (qx_needs_dequant) {
ggml_pipeline_allocate_descriptor_sets(ctx->device, to_fp16_vk_0, 1);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
if (qy_needs_dequant) {
ggml_pipeline_allocate_descriptor_sets(ctx->device, to_fp16_vk_1, y_non_contig ? 1 : ne12 * ne13);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
ggml_pipeline_allocate_descriptor_sets(ctx->device, dmmv, ne12 * ne13);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
if (x_non_contig) {
GGML_ASSERT(x_sz == ggml_vk_align_size(ggml_type_size(src0->type) * x_ne, ctx->device->properties.limits.minStorageBufferOffsetAlignment));
ggml_vk_cpy_to_contiguous(ctx, subctx, to_fp16_vk_0, src0, { d_Qx, qx_buf_offset, VK_WHOLE_SIZE }, { d_X, 0, VK_WHOLE_SIZE });
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
if (y_non_contig) {
GGML_ASSERT(y_sz == ggml_type_size(src1->type) * y_ne);
ggml_vk_cpy_to_contiguous(ctx, subctx, to_fp16_vk_1, src1, { d_Qy, qy_buf_offset, VK_WHOLE_SIZE }, { d_Y, 0, VK_WHOLE_SIZE });
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
uint32_t stride_batch_x = ne00*ne01;
uint32_t stride_batch_y = ne10*ne11;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
if (!ggml_vk_dim01_contiguous(src0) && !qx_needs_dequant) {
stride_batch_x = src0->nb[0] / ggml_type_size(src0->type);
}
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
if (!ggml_vk_dim01_contiguous(src1) && !qy_needs_dequant) {
stride_batch_y = src1->nb[0] / ggml_type_size(src1->type);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
const uint32_t max_groups_x = ctx->device->properties.limits.maxComputeWorkGroupCount[0];
uint32_t groups_x = ne01;
uint32_t groups_z = 1;
if (ne01 > max_groups_x) {
groups_z = 64;
groups_x /= groups_z;
}
// compute
const vk_mat_vec_push_constants pc = {
(uint32_t)ne00, (uint32_t)ne10, (uint32_t)ne10, (uint32_t)ne01,
stride_batch_x, stride_batch_y, (uint32_t)(ne20*ne21),
(uint32_t)ne02, (uint32_t)ne12, (uint32_t)r2, (uint32_t)r3,
};
ggml_vk_sync_buffers(subctx);
ggml_vk_dispatch_pipeline(ctx, subctx, dmmv,
{ { d_X, x_buf_offset, x_sz * ne02 * ne03 }, { d_Y, y_buf_offset, y_sz * ne12 * ne13 }, { d_D, d_buf_offset, d_sz * ne22 * ne23} },
sizeof(vk_mat_vec_push_constants), &pc, { groups_x, (uint32_t)(ne12 * ne13), groups_z });
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
static void ggml_vk_mul_mat_vec_p021_f16_f32(ggml_backend_vk_context * ctx, vk_context * subctx, const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst) {
VK_LOG_DEBUG("ggml_vk_mul_mat_p021_f16_f32(" << src0 << ", name=" << src0->name << ", type=" << src0->type << ", ne0=" << src0->ne[0] << ", ne1=" << src0->ne[1] << ", ne2=" << src0->ne[2] << ", ne3=" << src0->ne[3] << ", nb0=" << src0->nb[0] << ", nb1=" << src0->nb[1] << ", nb2=" << src0->nb[2] << ", nb3=" << src0->nb[3];
std::cerr << "), (" << src1 << ", name=" << src1->name << ", type=" << src1->type << ", ne0=" << src1->ne[0] << ", ne1=" << src1->ne[1] << ", ne2=" << src1->ne[2] << ", ne3=" << src1->ne[3] << ", nb0=" << src1->nb[0] << ", nb1=" << src1->nb[1] << ", nb2=" << src1->nb[2] << ", nb3=" << src1->nb[3];
std::cerr << "), (" << dst << ", name=" << dst->name << ", type=" << dst->type << ", ne0=" << dst->ne[0] << ", ne1=" << dst->ne[1] << ", ne2=" << dst->ne[2] << ", ne3=" << dst->ne[3] << ", nb0=" << dst->nb[0] << ", nb1=" << dst->nb[1] << ", nb2=" << dst->nb[2] << ", nb3=" << dst->nb[3] << "),)");
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
GGML_ASSERT(ggml_is_permuted(src0) && ggml_is_permuted(src1));
GGML_ASSERT(src0->nb[0] <= src0->nb[1] && src0->nb[2] <= src0->nb[3]); // NOLINT
GGML_ASSERT(src1->nb[0] <= src1->nb[1] && src1->nb[2] <= src1->nb[3]); // NOLINT
GGML_ASSERT(src0->type == GGML_TYPE_F16);
GGML_ASSERT(src1->type == GGML_TYPE_F32);
const uint64_t ne00 = src0->ne[0];
const uint64_t ne01 = src0->ne[1];
const uint64_t ne02 = src0->ne[2];
// const uint64_t ne03 = src0->ne[3];
const uint64_t ne10 = src1->ne[0];
const uint64_t ne11 = src1->ne[1];
const uint64_t ne12 = src1->ne[2];
// const uint64_t ne13 = src1->ne[3];
GGML_ASSERT(ne11 == 1);
ggml_tensor_extra_gpu * extra = (ggml_tensor_extra_gpu *) dst->extra;
ggml_tensor_extra_gpu * extra_src0 = (ggml_tensor_extra_gpu *) src0->extra;
ggml_tensor_extra_gpu * extra_src1 = (ggml_tensor_extra_gpu *) src1->extra;
vk_buffer d_Qy;
size_t qy_buf_offset = 0;
bool src1_uma = false;
if (ctx->device->uma) {
ggml_vk_host_get(ctx->device, src1->data, d_Qy, qy_buf_offset);
src1_uma = d_Qy != nullptr;
}
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
const uint64_t x_ne = ne00 * ne01 * ne02;
const uint64_t y_ne = ne10 * ne11 * ne12;
const uint64_t d_ne = ne01 * ne11 * ne12;
const uint64_t qx_sz = ggml_vk_align_size(ggml_type_size(src0->type) * x_ne / ggml_blck_size(src0->type), ctx->device->properties.limits.minStorageBufferOffsetAlignment);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
const uint64_t qy_sz = ggml_type_size(src1->type) * y_ne / ggml_blck_size(src1->type);
const uint64_t d_sz = sizeof(float) * d_ne;
vk_buffer d_D = extra->buffer_gpu.lock();
const uint64_t d_buf_offset = extra->offset + dst->view_offs;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
GGML_ASSERT(d_D != nullptr);
vk_buffer d_Qx = extra_src0->buffer_gpu.lock();
const uint64_t qx_buf_offset = extra_src0->offset + src0->view_offs;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
GGML_ASSERT(d_Qx != nullptr);
if (!src1_uma) {
d_Qy = extra_src1->buffer_gpu.lock();
qy_buf_offset = extra_src1->offset + src1->view_offs;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
GGML_ASSERT(d_Qx != nullptr);
}
// Allocate descriptor sets
ggml_pipeline_allocate_descriptor_sets(ctx->device, ctx->device->pipeline_mul_mat_vec_p021_f16_f32, 1);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
const uint64_t qy_buffer_offset = (qy_buf_offset / ctx->device->properties.limits.minStorageBufferOffsetAlignment) * ctx->device->properties.limits.minStorageBufferOffsetAlignment;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
const uint64_t qy_shader_offset = qy_buf_offset - qy_buffer_offset;
const uint64_t d_buffer_offset = (d_buf_offset / ctx->device->properties.limits.minStorageBufferOffsetAlignment) * ctx->device->properties.limits.minStorageBufferOffsetAlignment;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
const uint64_t d_shader_offset = d_buf_offset - d_buffer_offset;
// compute
const std::array<uint32_t, 6> pc = { (uint32_t)ne00, (uint32_t)ne01, (uint32_t)ne02, (uint32_t)ne12, (uint32_t)(qy_shader_offset / ggml_type_size(src1->type)), (uint32_t)(d_shader_offset / ggml_type_size(dst->type)) };
ggml_vk_sync_buffers(subctx);
ggml_vk_dispatch_pipeline(ctx, subctx, ctx->device->pipeline_mul_mat_vec_p021_f16_f32, { { d_Qx, qx_buf_offset, qx_sz }, { d_Qy, qy_buffer_offset, qy_sz + qy_shader_offset }, { d_D, d_buffer_offset, d_sz + d_shader_offset } }, 6 * sizeof(uint32_t), &pc, { 1, (uint32_t)ne01, (uint32_t)ne12 });
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
static void ggml_vk_mul_mat_vec_nc_f16_f32(ggml_backend_vk_context * ctx, vk_context * subctx, const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst) {
VK_LOG_DEBUG("ggml_vk_mul_mat_nc_f16_f32((" << src0 << ", name=" << src0->name << ", type=" << src0->type << ", ne0=" << src0->ne[0] << ", ne1=" << src0->ne[1] << ", ne2=" << src0->ne[2] << ", ne3=" << src0->ne[3] << ", nb0=" << src0->nb[0] << ", nb1=" << src0->nb[1] << ", nb2=" << src0->nb[2] << ", nb3=" << src0->nb[3];
std::cerr << "), (" << src1 << ", name=" << src1->name << ", type=" << src1->type << ", ne0=" << src1->ne[0] << ", ne1=" << src1->ne[1] << ", ne2=" << src1->ne[2] << ", ne3=" << src1->ne[3] << ", nb0=" << src1->nb[0] << ", nb1=" << src1->nb[1] << ", nb2=" << src1->nb[2] << ", nb3=" << src1->nb[3];
std::cerr << "), (" << dst << ", name=" << dst->name << ", type=" << dst->type << ", ne0=" << dst->ne[0] << ", ne1=" << dst->ne[1] << ", ne2=" << dst->ne[2] << ", ne3=" << dst->ne[3] << ", nb0=" << dst->nb[0] << ", nb1=" << dst->nb[1] << ", nb2=" << dst->nb[2] << ", nb3=" << dst->nb[3] << "),)");
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
GGML_ASSERT(!ggml_is_transposed(src0));
GGML_ASSERT(!ggml_is_transposed(src1));
GGML_ASSERT(!ggml_is_permuted(src0));
GGML_ASSERT(src0->type == GGML_TYPE_F16);
GGML_ASSERT(src1->type == GGML_TYPE_F32);
const uint64_t ne00 = src0->ne[0];
const uint64_t ne01 = src0->ne[1];
const uint64_t ne02 = src0->ne[2];
// const uint64_t ne03 = src0->ne[3];
const uint64_t nb01 = src0->nb[1];
const uint64_t nb02 = src0->nb[2];
// const uint64_t ne10 = src1->ne[0];
const uint64_t ne11 = src1->ne[1];
const uint64_t ne12 = src1->ne[2];
// const uint64_t ne13 = src1->ne[3];
GGML_ASSERT(ne11 == 1);
ggml_tensor_extra_gpu * extra = (ggml_tensor_extra_gpu *) dst->extra;
ggml_tensor_extra_gpu * extra_src0 = (ggml_tensor_extra_gpu *) src0->extra;
ggml_tensor_extra_gpu * extra_src1 = (ggml_tensor_extra_gpu *) src1->extra;
vk_buffer d_Qy = nullptr;
size_t qy_buf_offset = 0;
bool src1_uma = false;
if (ctx->device->uma) {
ggml_vk_host_get(ctx->device, src1->data, d_Qy, qy_buf_offset);
src1_uma = d_Qy != nullptr;
}
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
const uint64_t d_ne = ne01 * ne11 * ne12;
const uint32_t row_stride_x = nb01 / sizeof(ggml_fp16_t);
const uint32_t channel_stride_x = nb02 / sizeof(ggml_fp16_t);
const uint64_t qx_sz = ggml_nbytes(src0);
const uint64_t qy_sz = ggml_nbytes(src1);
const uint64_t d_sz = sizeof(float) * d_ne;
vk_buffer d_D = extra->buffer_gpu.lock();
const uint64_t d_buf_offset = extra->offset + dst->view_offs;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
GGML_ASSERT(d_D != nullptr);
vk_buffer d_Qx = extra_src0->buffer_gpu.lock();
const uint64_t qx_buf_offset = extra_src0->offset + src0->view_offs;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
GGML_ASSERT(d_Qx != nullptr);
if (!src1_uma) {
d_Qy = extra_src1->buffer_gpu.lock();
qy_buf_offset = extra_src1->offset + src1->view_offs;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
GGML_ASSERT(d_Qx != nullptr);
}
// Allocate descriptor sets
ggml_pipeline_allocate_descriptor_sets(ctx->device, ctx->device->pipeline_mul_mat_vec_nc_f16_f32, 1);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
const uint64_t qy_buffer_offset = (qy_buf_offset / ctx->device->properties.limits.minStorageBufferOffsetAlignment) * ctx->device->properties.limits.minStorageBufferOffsetAlignment;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
const uint64_t qy_shader_offset = qy_buf_offset - qy_buffer_offset;
const uint64_t d_buffer_offset = (d_buf_offset / ctx->device->properties.limits.minStorageBufferOffsetAlignment) * ctx->device->properties.limits.minStorageBufferOffsetAlignment;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
const uint64_t d_shader_offset = d_buf_offset - d_buffer_offset;
// compute
const std::array<uint32_t, 7> pc = { (uint32_t)ne00, (uint32_t)ne01, row_stride_x, channel_stride_x, (uint32_t)(ne12 / ne02), (uint32_t)(qy_shader_offset / ggml_type_size(src1->type)), (uint32_t)(d_shader_offset / ggml_type_size(dst->type)) };
ggml_vk_sync_buffers(subctx);
ggml_vk_dispatch_pipeline(ctx, subctx, ctx->device->pipeline_mul_mat_vec_nc_f16_f32, { { d_Qx, qx_buf_offset, qx_sz }, { d_Qy, qy_buffer_offset, qy_sz + qy_shader_offset }, { d_D, d_buffer_offset, d_sz + d_shader_offset } }, 7 * sizeof(uint32_t), &pc, { 1, (uint32_t)ne01, (uint32_t)ne12 });
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
static void ggml_vk_mul_mat(ggml_backend_vk_context * ctx, vk_context * subctx, const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst) {
VK_LOG_DEBUG("ggml_vk_mul_mat(" << src0 << ", " << src1 << ", " << dst << ")");
if (src0->type == GGML_TYPE_F16 && ggml_is_permuted(src0) && ggml_is_permuted(src1) && dst->ne[1] == 1) {
ggml_vk_mul_mat_vec_p021_f16_f32(ctx, subctx, src0, src1, dst);
} else if (src0->type == GGML_TYPE_F16 && !ggml_is_contiguous(src0) && !ggml_is_transposed(src1) && dst->ne[1] == 1) {
ggml_vk_mul_mat_vec_nc_f16_f32(ctx, subctx, src0, src1, dst);
} else if (dst->ne[1] == 1 && (src0->type == GGML_TYPE_F32 || src0->type == GGML_TYPE_F16 || ggml_is_quantized(src0->type))) {
ggml_vk_mul_mat_vec_q_f16(ctx, subctx, src0, src1, dst);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
} else {
ggml_vk_mul_mat_q_f16(ctx, subctx, src0, src1, dst);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
}
static void ggml_vk_mul_mat_id_q_f16(ggml_backend_vk_context * ctx, vk_context * subctx, const ggml_tensor * src0, const ggml_tensor * src1, const ggml_tensor * ids, ggml_tensor * dst) {
VK_LOG_DEBUG("ggml_vk_mul_mat_id_q_f16((" << src0 << ", name=" << src0->name << ", type=" << src0->type << ", ne0=" << src0->ne[0] << ", ne1=" << src0->ne[1] << ", ne2=" << src0->ne[2] << ", ne3=" << src0->ne[3] << ", nb0=" << src0->nb[0] << ", nb1=" << src0->nb[1] << ", nb2=" << src0->nb[2] << ", nb3=" << src0->nb[3];
std::cerr << "), (" << src1 << ", name=" << src1->name << ", type=" << src1->type << ", ne0=" << src1->ne[0] << ", ne1=" << src1->ne[1] << ", ne2=" << src1->ne[2] << ", ne3=" << src1->ne[3] << ", nb0=" << src1->nb[0] << ", nb1=" << src1->nb[1] << ", nb2=" << src1->nb[2] << ", nb3=" << src1->nb[3];
std::cerr << "), (" << ids << ", name=" << ids->name << ", type=" << ids->type << ", ne0=" << ids->ne[0] << ", ne1=" << ids->ne[1] << ", ne2=" << ids->ne[2] << ", ne3=" << ids->ne[3] << ", nb0=" << ids->nb[0] << ", nb1=" << ids->nb[1] << ", nb2=" << ids->nb[2] << ", nb3=" << ids->nb[3];
std::cerr << "), (" << dst << ", name=" << dst->name << ", type=" << dst->type << ", ne0=" << dst->ne[0] << ", ne1=" << dst->ne[1] << ", ne2=" << dst->ne[2] << ", ne3=" << dst->ne[3] << ", nb0=" << dst->nb[0] << ", nb1=" << dst->nb[1] << ", nb2=" << dst->nb[2] << ", nb3=" << dst->nb[3] << "),)");
GGML_ASSERT(ggml_vk_dim01_contiguous(src1) || src1->type == GGML_TYPE_F32 || src1->type == GGML_TYPE_F16); // NOLINT
GGML_ASSERT(ids->type == GGML_TYPE_I32);
const uint64_t ne00 = src0->ne[0];
const uint64_t ne01 = src0->ne[1];
const uint64_t ne02 = src0->ne[2];
const uint64_t ne03 = src0->ne[3];
const uint64_t ne10 = src1->ne[0];
const uint64_t ne11 = src1->ne[1];
const uint64_t ne12 = src1->ne[2];
const uint64_t ne13 = src1->ne[3];
const uint64_t nei0 = ids->ne[0];
const uint64_t nei1 = ids->ne[1];
GGML_ASSERT(nei0 * nei1 <= 2048);
const uint32_t nbi1 = ids->nb[1];
const uint32_t nbi2 = ids->nb[2];
const uint64_t ne20 = dst->ne[0];
const uint64_t ne21 = dst->ne[1];
const uint64_t ne22 = dst->ne[2];
const uint64_t ne23 = dst->ne[3];
const uint64_t n_as = ne02;
GGML_ASSERT(n_as <= 8);
ggml_tensor_extra_gpu * extra = (ggml_tensor_extra_gpu *) dst->extra;
ggml_tensor_extra_gpu * extra_src0 = (ggml_tensor_extra_gpu *) src0->extra;
ggml_tensor_extra_gpu * extra_src1 = (ggml_tensor_extra_gpu *) src1->extra;
ggml_tensor_extra_gpu * extra_ids = (ggml_tensor_extra_gpu *) ids->extra;
vk_buffer d_Qx;
size_t qx_buf_offset = 0;
vk_buffer d_Qy;
size_t qy_buf_offset = 0;
vk_buffer d_ids;
size_t ids_buf_offset = 0;
bool src0_uma = false;
bool src1_uma = false;
bool ids_uma = false;
if (ctx->device->uma) {
ggml_vk_host_get(ctx->device, src0->data, d_Qx, qx_buf_offset);
ggml_vk_host_get(ctx->device, src1->data, d_Qy, qy_buf_offset);
ggml_vk_host_get(ctx->device, ids->data, d_ids, ids_buf_offset);
src0_uma = d_Qx != nullptr;
src1_uma = d_Qy != nullptr;
ids_uma = d_ids != nullptr;
}
const bool x_non_contig = !ggml_vk_dim01_contiguous(src0);
const bool y_non_contig = !ggml_vk_dim01_contiguous(src1);
const bool y_f32_kernel = src1->type == GGML_TYPE_F32 && !y_non_contig;
vk_matmul_pipeline mmp = ggml_vk_get_mul_mat_mat_id_pipeline(ctx, src0->type, y_non_contig ? GGML_TYPE_F16 : src1->type);
const bool qx_needs_dequant = mmp == nullptr || x_non_contig;
const bool qy_needs_dequant = (src1->type != GGML_TYPE_F16 && !y_f32_kernel) || y_non_contig;
if (mmp == nullptr) {
GGML_ASSERT(false);
}
// Not implemented
GGML_ASSERT(y_non_contig || !qy_needs_dequant); // NOLINT
const uint64_t x_ne = ne01 * ne00;
const uint64_t y_ne = ne11 * ne10;
const uint64_t d_ne = ne21 * ne20;
const uint32_t kpad = ggml_vk_align_size(ne10, ggml_vk_guess_matmul_pipeline_align(ctx, mmp, ne01, nei1));
const bool aligned = ne10 == kpad && ne01 > 8 && nei1 > 8;
vk_pipeline pipeline = ggml_vk_guess_matmul_pipeline(ctx, mmp, ne01, nei1, aligned);
const uint64_t qx_sz = ggml_type_size(src0->type) * x_ne / ggml_blck_size(src0->type);
const uint64_t qy_sz = ggml_type_size(src1->type) * y_ne / ggml_blck_size(src1->type);
const uint64_t x_sz = !qx_needs_dequant ? qx_sz : sizeof(ggml_fp16_t) * x_ne;
const uint64_t y_sz = y_f32_kernel ? sizeof(float) * y_ne : sizeof(ggml_fp16_t) * y_ne;
const uint64_t ids_sz = nbi2;
const uint64_t d_sz = sizeof(float) * d_ne;
vk_buffer d_D = extra->buffer_gpu.lock();
const uint64_t d_buf_offset = extra->offset + dst->view_offs;
GGML_ASSERT(d_D != nullptr);
vk_buffer d_X;
uint64_t x_buf_offset = 0;
vk_buffer d_Y;
uint64_t y_buf_offset = 0;
if (!src0_uma) {
d_Qx = extra_src0->buffer_gpu.lock();
qx_buf_offset = extra_src0->offset + src0->view_offs;
GGML_ASSERT(d_Qx != nullptr);
}
if (!src1_uma) {
d_Qy = extra_src1->buffer_gpu.lock();
qy_buf_offset = extra_src1->offset + src1->view_offs;
GGML_ASSERT(d_Qy != nullptr);
}
if (!ids_uma) {
d_ids = extra_ids->buffer_gpu.lock();
ids_buf_offset = extra_ids->offset + ids->view_offs;
GGML_ASSERT(d_ids != nullptr);
}
if (qx_needs_dequant) {
d_X = ctx->prealloc_x;
GGML_ASSERT(d_X->size >= x_sz * ne02 * ne03);
} else {
d_X = d_Qx;
x_buf_offset = qx_buf_offset;
GGML_ASSERT(qx_sz == x_sz);
}
if (qy_needs_dequant) {
d_Y = ctx->prealloc_y;
GGML_ASSERT(d_Y->size >= y_sz * ne02 * ne03);
} else {
d_Y = d_Qy;
y_buf_offset = qy_buf_offset;
GGML_ASSERT(qy_sz == y_sz);
}
vk_pipeline to_fp16_vk_0 = nullptr;
vk_pipeline to_fp16_vk_1 = nullptr;
if (x_non_contig) {
to_fp16_vk_0 = ggml_vk_get_cpy_pipeline(ctx, src0->type, GGML_TYPE_F16);
} else {
to_fp16_vk_0 = ggml_vk_get_to_fp16(ctx, src0->type);
}
if (y_non_contig) {
to_fp16_vk_1 = ggml_vk_get_cpy_pipeline(ctx, src1->type, GGML_TYPE_F16);
} else {
to_fp16_vk_1 = ggml_vk_get_to_fp16(ctx, src1->type);
}
GGML_ASSERT(!qx_needs_dequant || to_fp16_vk_0 != nullptr); // NOLINT
GGML_ASSERT(!qy_needs_dequant || to_fp16_vk_1 != nullptr); // NOLINT
// Allocate descriptor sets
ggml_pipeline_allocate_descriptor_sets(ctx->device, pipeline, 1);
if (qx_needs_dequant) {
ggml_pipeline_allocate_descriptor_sets(ctx->device, to_fp16_vk_0, 1);
}
if (qy_needs_dequant) {
ggml_pipeline_allocate_descriptor_sets(ctx->device, to_fp16_vk_1, 1);
}
if (x_non_contig) {
ggml_vk_cpy_to_contiguous(ctx, subctx, to_fp16_vk_0, src0, { d_Qx, qx_buf_offset, VK_WHOLE_SIZE }, { d_X, 0, VK_WHOLE_SIZE });
} else if (qx_needs_dequant) {
const std::vector<uint32_t> pc = { (uint32_t)ne01, (uint32_t)ne10, (uint32_t)ne10, (uint32_t)ne10, (uint32_t)(ggml_nelements(src0)) };
ggml_vk_sync_buffers(subctx);
ggml_vk_dispatch_pipeline(ctx, subctx, to_fp16_vk_0, { { d_Qx, qx_buf_offset, qx_sz * ne02 * ne03 }, { d_X, 0, x_sz * ne02 * ne03 } }, pc.size() * sizeof(uint32_t), pc.data(), { (uint32_t)(x_ne * ne02 * ne03), 1, 1});
}
if (y_non_contig) {
ggml_vk_cpy_to_contiguous(ctx, subctx, to_fp16_vk_1, src1, { d_Qy, qy_buf_offset, VK_WHOLE_SIZE }, { d_Y, 0, VK_WHOLE_SIZE });
}
uint32_t stride_batch_x = ne00*ne01;
uint32_t stride_batch_y = ne10*ne11;
if (!ggml_vk_dim01_contiguous(src0) && !qx_needs_dequant) {
stride_batch_x = src0->nb[0] / ggml_type_size(src0->type);
}
if (!ggml_vk_dim01_contiguous(src1) && !qy_needs_dequant) {
stride_batch_y = src1->nb[0] / ggml_type_size(src1->type);
}
// compute
ggml_vk_matmul_id(
ctx, subctx, pipeline,
{ d_X, x_buf_offset, x_sz * ne02 * ne03 }, { d_Y, y_buf_offset, y_sz * ne12 * ne13 },
{ d_D, d_buf_offset, d_sz * ne22 * ne23 }, { d_ids, ids_buf_offset, ids_sz },
ne01, ne21, ne10, ne10, ne10, ne01,
stride_batch_x, stride_batch_y, ne20*ne21,
n_as, nei0, nei1, nbi1 / ggml_type_size(ids->type), ne11
); // NOLINT
}
static void ggml_vk_mul_mat_vec_id_q_f16(ggml_backend_vk_context * ctx, vk_context * subctx, const ggml_tensor * src0, const ggml_tensor * src1, const ggml_tensor * ids, ggml_tensor * dst) {
VK_LOG_DEBUG("ggml_vk_mul_mat_vec_id_q_f16((" << src0 << ", name=" << src0->name << ", type=" << src0->type << ", ne0=" << src0->ne[0] << ", ne1=" << src0->ne[1] << ", ne2=" << src0->ne[2] << ", ne3=" << src0->ne[3] << ", nb0=" << src0->nb[0] << ", nb1=" << src0->nb[1] << ", nb2=" << src0->nb[2] << ", nb3=" << src0->nb[3];
std::cerr << "), (" << src1 << ", name=" << src1->name << ", type=" << src1->type << ", ne0=" << src1->ne[0] << ", ne1=" << src1->ne[1] << ", ne2=" << src1->ne[2] << ", ne3=" << src1->ne[3] << ", nb0=" << src1->nb[0] << ", nb1=" << src1->nb[1] << ", nb2=" << src1->nb[2] << ", nb3=" << src1->nb[3];
std::cerr << "), (" << ids << ", name=" << ids->name << ", type=" << ids->type << ", ne0=" << ids->ne[0] << ", ne1=" << ids->ne[1] << ", ne2=" << ids->ne[2] << ", ne3=" << ids->ne[3] << ", nb0=" << ids->nb[0] << ", nb1=" << ids->nb[1] << ", nb2=" << ids->nb[2] << ", nb3=" << ids->nb[3];
std::cerr << "), (" << dst << ", name=" << dst->name << ", type=" << dst->type << ", ne0=" << dst->ne[0] << ", ne1=" << dst->ne[1] << ", ne2=" << dst->ne[2] << ", ne3=" << dst->ne[3] << ", nb0=" << dst->nb[0] << ", nb1=" << dst->nb[1] << ", nb2=" << dst->nb[2] << ", nb3=" << dst->nb[3] << "),)");
GGML_ASSERT(ggml_vk_dim01_contiguous(src0) || src0->type == GGML_TYPE_F32 || src0->type == GGML_TYPE_F16); // NOLINT
GGML_ASSERT(ggml_vk_dim01_contiguous(src1) || src1->type == GGML_TYPE_F32 || src1->type == GGML_TYPE_F16); // NOLINT
GGML_ASSERT(ids->type == GGML_TYPE_I32);
const uint64_t ne00 = src0->ne[0];
const uint64_t ne01 = src0->ne[1];
const uint64_t ne02 = src0->ne[2];
const uint64_t ne03 = src0->ne[3];
const uint64_t ne10 = src1->ne[0];
const uint64_t ne11 = src1->ne[1];
const uint64_t ne12 = src1->ne[2];
const uint64_t ne13 = src1->ne[3];
const uint64_t nei0 = ids->ne[0];
const uint64_t nei1 = ids->ne[1];
const uint64_t nbi2 = ids->nb[2];
GGML_ASSERT(nei1 == 1);
const uint64_t ne20 = dst->ne[0];
const uint64_t ne21 = dst->ne[1];
const uint64_t ne22 = dst->ne[2];
const uint64_t ne23 = dst->ne[3];
ggml_tensor_extra_gpu * extra = (ggml_tensor_extra_gpu *) dst->extra;
ggml_tensor_extra_gpu * extra_src0 = (ggml_tensor_extra_gpu *) src0->extra;
ggml_tensor_extra_gpu * extra_src1 = (ggml_tensor_extra_gpu *) src1->extra;
ggml_tensor_extra_gpu * extra_ids = (ggml_tensor_extra_gpu *) ids->extra;
vk_buffer d_Qx;
size_t qx_buf_offset = 0;
vk_buffer d_Qy;
size_t qy_buf_offset = 0;
vk_buffer d_ids;
size_t ids_buf_offset = 0;
bool src0_uma = false;
bool src1_uma = false;
bool ids_uma = false;
if (ctx->device->uma) {
ggml_vk_host_get(ctx->device, src0->data, d_Qx, qx_buf_offset);
ggml_vk_host_get(ctx->device, src1->data, d_Qy, qy_buf_offset);
ggml_vk_host_get(ctx->device, ids->data, d_ids, ids_buf_offset);
src0_uma = d_Qx != nullptr;
src1_uma = d_Qy != nullptr;
ids_uma = d_ids != nullptr;
}
const bool x_non_contig = !ggml_vk_dim01_contiguous(src0);
const bool y_non_contig = !ggml_vk_dim01_contiguous(src1);
const bool f16_f32_kernel = src1->type == GGML_TYPE_F32;
const bool qx_needs_dequant = x_non_contig;
const bool qy_needs_dequant = (src1->type != GGML_TYPE_F16 && !f16_f32_kernel) || y_non_contig;
// Not implemented
GGML_ASSERT(y_non_contig || !qy_needs_dequant); // NOLINT
const uint64_t x_ne = ne01 * ne00;
const uint64_t y_ne = ne11 * ne10;
const uint64_t d_ne = ne21 * ne20;
const uint64_t qx_sz = ggml_vk_align_size(ggml_type_size(src0->type) * x_ne / ggml_blck_size(src0->type), ctx->device->properties.limits.minStorageBufferOffsetAlignment);
const uint64_t qy_sz = ggml_type_size(src1->type) * y_ne / ggml_blck_size(src1->type);
const uint64_t x_sz = x_non_contig ? ggml_vk_align_size(ggml_type_size(src0->type) * x_ne, ctx->device->properties.limits.minStorageBufferOffsetAlignment) : qx_sz;
const uint64_t y_sz = f16_f32_kernel ? sizeof(float) * y_ne : sizeof(ggml_fp16_t) * y_ne;
const uint64_t ids_sz = nbi2;
const uint64_t d_sz = sizeof(float) * d_ne;
vk_buffer d_D = extra->buffer_gpu.lock();
const uint64_t d_buf_offset = extra->offset + dst->view_offs;
GGML_ASSERT(d_D != nullptr);
vk_buffer d_X;
uint64_t x_buf_offset = 0;
vk_buffer d_Y;
uint64_t y_buf_offset = 0;
if(!src0_uma) {
d_Qx = extra_src0->buffer_gpu.lock();
qx_buf_offset = extra_src0->offset + src0->view_offs;
GGML_ASSERT(d_Qx != nullptr);
}
if(!src1_uma) {
d_Qy = extra_src1->buffer_gpu.lock();
qy_buf_offset = extra_src1->offset + src1->view_offs;
GGML_ASSERT(d_Qy != nullptr);
}
if(!ids_uma) {
d_ids = extra_ids->buffer_gpu.lock();
ids_buf_offset = extra_ids->offset + ids->view_offs;
GGML_ASSERT(d_ids != nullptr);
}
if (qx_needs_dequant) {
d_X = ctx->prealloc_x;
} else {
d_X = d_Qx;
x_buf_offset = qx_buf_offset;
GGML_ASSERT(qx_sz == x_sz);
}
if (qy_needs_dequant) {
d_Y = ctx->prealloc_y;
} else {
d_Y = d_Qy;
y_buf_offset = qy_buf_offset;
GGML_ASSERT(qy_sz == y_sz);
}
vk_pipeline to_fp16_vk_0 = nullptr;
vk_pipeline to_fp16_vk_1 = nullptr;
if (x_non_contig) {
to_fp16_vk_0 = ggml_vk_get_cpy_pipeline(ctx, src0->type, src0->type);
}
if (y_non_contig) {
to_fp16_vk_1 = ggml_vk_get_cpy_pipeline(ctx, src1->type, src1->type);
} else {
to_fp16_vk_1 = ggml_vk_get_to_fp16(ctx, src1->type);
}
vk_pipeline dmmv = ggml_vk_get_dequantize_mul_mat_vec_id(ctx, src0->type, src1->type);
GGML_ASSERT(!qx_needs_dequant || to_fp16_vk_0 != nullptr); // NOLINT
GGML_ASSERT(!qy_needs_dequant || to_fp16_vk_1 != nullptr); // NOLINT
GGML_ASSERT(dmmv != nullptr);
// Allocate descriptor sets
if (qx_needs_dequant) {
ggml_pipeline_allocate_descriptor_sets(ctx->device, to_fp16_vk_0, 1);
}
if (qy_needs_dequant) {
ggml_pipeline_allocate_descriptor_sets(ctx->device, to_fp16_vk_1, y_non_contig ? 1 : ne12 * ne13);
}
ggml_pipeline_allocate_descriptor_sets(ctx->device, dmmv, ne12 * ne13);
if (x_non_contig) {
GGML_ASSERT(x_sz == ggml_vk_align_size(ggml_type_size(src0->type) * x_ne, ctx->device->properties.limits.minStorageBufferOffsetAlignment));
ggml_vk_cpy_to_contiguous(ctx, subctx, to_fp16_vk_0, src0, { d_Qx, qx_buf_offset, VK_WHOLE_SIZE }, { d_X, 0, VK_WHOLE_SIZE });
}
if (y_non_contig) {
GGML_ASSERT(y_sz == ggml_type_size(src1->type) * y_ne);
ggml_vk_cpy_to_contiguous(ctx, subctx, to_fp16_vk_1, src1, { d_Qy, qy_buf_offset, VK_WHOLE_SIZE }, { d_Y, 0, VK_WHOLE_SIZE });
}
uint32_t stride_batch_y = ne10*ne11;
if (!ggml_vk_dim01_contiguous(src1) && !qy_needs_dequant) {
stride_batch_y = src1->nb[0] / ggml_type_size(src1->type);
}
const uint32_t max_groups_x = ctx->device->properties.limits.maxComputeWorkGroupCount[0];
uint32_t groups_x = ne01;
uint32_t groups_z = 1;
if (ne01 > max_groups_x) {
groups_z = 64;
groups_x /= groups_z;
}
// compute
const vk_mat_vec_id_push_constants pc = {
(uint32_t)ne00, (uint32_t)ne10, (uint32_t)ne10, (uint32_t)ne01,
(uint32_t)x_ne, stride_batch_y, (uint32_t)(ne20*ne21),
(uint32_t)nei0, (uint32_t)ne11,
};
ggml_vk_sync_buffers(subctx);
ggml_vk_dispatch_pipeline(ctx, subctx, dmmv,
{ { d_X, x_buf_offset, x_sz * ne02 * ne03 }, { d_Y, y_buf_offset, y_sz * ne12 * ne13 }, { d_D, d_buf_offset, d_sz * ne22 * ne23}, { d_ids, ids_buf_offset, ids_sz } },
sizeof(vk_mat_vec_id_push_constants), &pc, { groups_x, (uint32_t)nei0, groups_z });
}
static void ggml_vk_mul_mat_id(ggml_backend_vk_context * ctx, vk_context * subctx, const ggml_tensor * src0, const ggml_tensor * src1, const ggml_tensor * src2, ggml_tensor * dst) {
VK_LOG_DEBUG("ggml_vk_mul_mat_id(" << src0 << ", " << src1 << ", " << src2 << ", " << dst << ")");
if (src2->ne[1] == 1 && (src0->type == GGML_TYPE_F32 || src0->type == GGML_TYPE_F16 || ggml_is_quantized(src0->type))) {
ggml_vk_mul_mat_vec_id_q_f16(ctx, subctx, src0, src1, src2, dst);
} else {
ggml_vk_mul_mat_id_q_f16(ctx, subctx, src0, src1, src2, dst);
}
}
static void ggml_vk_op_repeat(ggml_backend_vk_context * ctx, vk_context * subctx, const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst) {
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
// guaranteed to be an integer due to the check in ggml_can_repeat
const uint64_t ne0 = dst->ne[0];
const uint64_t ne1 = dst->ne[1];
const uint64_t ne2 = dst->ne[2];
const uint64_t ne3 = dst->ne[3];
const uint64_t ne00 = src0->ne[0];
const uint64_t ne01 = src0->ne[1];
const uint64_t ne02 = src0->ne[2];
const uint64_t ne03 = src0->ne[3];
const uint64_t nb0 = dst->nb[0];
const uint64_t nb1 = dst->nb[1];
const uint64_t nb2 = dst->nb[2];
const uint64_t nb3 = dst->nb[3];
const uint64_t nb00 = src0->nb[0];
const uint64_t nb01 = src0->nb[1];
const uint64_t nb02 = src0->nb[2];
const uint64_t nb03 = src0->nb[3];
const uint64_t nr0 = ne0/ne00;
const uint64_t nr1 = ne1/ne01;
const uint64_t nr2 = ne2/ne02;
const uint64_t nr3 = ne3/ne03;
// TODO: support for transposed / permuted tensors
GGML_ASSERT(nb0 == sizeof(float));
GGML_ASSERT(nb00 == sizeof(float));
ggml_tensor_extra_gpu * extra = (ggml_tensor_extra_gpu *) dst->extra;
ggml_tensor_extra_gpu * extra_src0 = (ggml_tensor_extra_gpu *) src0->extra;
const vk_buffer src_buf = extra_src0->buffer_gpu.lock();
const uint64_t src_offset = extra_src0->offset + src0->view_offs;
vk_buffer dst_buf = extra->buffer_gpu.lock();
const uint64_t dst_offset = extra->offset + dst->view_offs;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
std::vector<vk::BufferCopy> copies;
for (uint64_t i3 = 0; i3 < nr3; i3++) {
for (uint64_t k3 = 0; k3 < ne03; k3++) {
for (uint64_t i2 = 0; i2 < nr2; i2++) {
for (uint64_t k2 = 0; k2 < ne02; k2++) {
for (uint64_t i1 = 0; i1 < nr1; i1++) {
for (uint64_t k1 = 0; k1 < ne01; k1++) {
for (uint64_t i0 = 0; i0 < nr0; i0++) {
copies.push_back({
src_offset + (i3*ne03 + k3)*nb3 + (i2*ne02 + k2)*nb2 + (i1*ne01 + k1)*nb1 + (i0*ne00)*nb0,
dst_offset + ( k3)*nb03 + ( k2)*nb02 + ( k1)*nb01,
ne00*nb0,
});
}
}
}
}
}
}
}
ggml_vk_sync_buffers(subctx);
subctx->s->buffer.copyBuffer(src_buf->buffer, dst_buf->buffer, copies);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
GGML_UNUSED(ctx);
GGML_UNUSED(src1);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
static vk_pipeline ggml_vk_op_get_pipeline(ggml_backend_vk_context * ctx, const ggml_tensor * src0, const ggml_tensor * src1, const ggml_tensor * src2, ggml_tensor * dst, ggml_op op) {
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
switch (op) {
case GGML_OP_ADD:
if (src0->type == GGML_TYPE_F32 && src1->type == GGML_TYPE_F32 && dst->type == GGML_TYPE_F32) {
return ctx->device->pipeline_add_f32;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
return nullptr;
case GGML_OP_GET_ROWS:
GGML_ASSERT(src1->type == GGML_TYPE_I32);
if (dst->type == GGML_TYPE_F16) {
return ctx->device->pipeline_get_rows[src0->type];
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
if (dst->type == GGML_TYPE_F32) {
return ctx->device->pipeline_get_rows_f32[src0->type];
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
return nullptr;
case GGML_OP_MUL:
if (src0->type == GGML_TYPE_F32 && src1->type == GGML_TYPE_F32 && dst->type == GGML_TYPE_F32) {
return ctx->device->pipeline_mul_f32;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
return nullptr;
case GGML_OP_DIV:
if (src0->type == GGML_TYPE_F32 && src1->type == GGML_TYPE_F32 && dst->type == GGML_TYPE_F32) {
return ctx->device->pipeline_div_f32;
}
return nullptr;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
case GGML_OP_SCALE:
if (src0->type == GGML_TYPE_F32 && dst->type == GGML_TYPE_F32) {
return ctx->device->pipeline_scale_f32;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
return nullptr;
case GGML_OP_SQR:
if (src0->type == GGML_TYPE_F32 && dst->type == GGML_TYPE_F32) {
return ctx->device->pipeline_sqr_f32;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
return nullptr;
case GGML_OP_CLAMP:
if (src0->type == GGML_TYPE_F32 && dst->type == GGML_TYPE_F32) {
return ctx->device->pipeline_clamp_f32;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
return nullptr;
case GGML_OP_CPY:
case GGML_OP_CONT:
case GGML_OP_DUP:
return ggml_vk_get_cpy_pipeline(ctx, src0->type, dst->type);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
case GGML_OP_NORM:
if (src0->type == GGML_TYPE_F32 && dst->type == GGML_TYPE_F32) {
return ctx->device->pipeline_norm_f32;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
return nullptr;
case GGML_OP_RMS_NORM:
if (src0->type == GGML_TYPE_F32 && dst->type == GGML_TYPE_F32) {
return ctx->device->pipeline_rms_norm_f32;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
return nullptr;
case GGML_OP_UNARY:
switch (ggml_get_unary_op(dst)) {
case GGML_UNARY_OP_SILU:
if (src0->type == GGML_TYPE_F32 && dst->type == GGML_TYPE_F32) {
return ctx->device->pipeline_silu_f32;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
break;
case GGML_UNARY_OP_GELU:
if (src0->type == GGML_TYPE_F32 && dst->type == GGML_TYPE_F32) {
return ctx->device->pipeline_gelu_f32;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
break;
case GGML_UNARY_OP_RELU:
if (src0->type == GGML_TYPE_F32 && dst->type == GGML_TYPE_F32) {
return ctx->device->pipeline_relu_f32;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
break;
default:
break;
}
return nullptr;
case GGML_OP_DIAG_MASK_INF:
if (src0->type == GGML_TYPE_F32 && dst->type == GGML_TYPE_F32) {
return ctx->device->pipeline_diag_mask_inf_f32;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
return nullptr;
case GGML_OP_SOFT_MAX:
GGML_ASSERT(!src1 || src1->type == GGML_TYPE_F32 || src1->type == GGML_TYPE_F16);
ggml : add Flash Attention (#5021) * ggml : add ggml_flash_attn_ext API * ggml : fix GQA support in ggml_flash_attn_ext * ggml : online attention (CPU) * metal : initial implementation * metal : f16 precision * metal : reduce branches * metal : specialize for head size * wip : 8 rows per simd group * wip : 4 rows per simd group * wip : template for rows per warp * metal : parallelize across KV size * metal : parallel reduce across heads * metal : efficient flash_attn_f16 implementation * metal : avoid redundant loads of the attention * metal : scale and mask in matrix form * metal : fix comment * llama : avoid ggml_cast, use F32 query * metal : add parallel reduce version (disabled) * metal : move output into local memory + optimize - the result from each simdgroup now stays in the registers - significantly reduced SRAM usage - more efficient skipping of -INF blocks - avoid simdgroup barrier in hot loop - add comments * metal : add tests, fix scaling, support C > 32 * metal : improve precision * ggml : fix f16 mad * metal : minor * metal : support Q > 8 * tests : add ATTN tests * metal : disable buffer allocation logs * tests : more * metal : faster inner loop for C == 32 * metal : fix array initialization * tests : ifdef * ggml : switch to padded F16 mask for ggml_soft_max, ggml_flash_attn_ext * ggml : fix ggml_soft_max mask requirement * cuda : fix soft_max to use correct mask size * cuda : add flash_attn kernel (wip) * metal : optimize softmax for C > 32 * metal : optimize softmax * tests : minor fix * cuda : avoid zeroing fragments * tests : update dims * cuda : fix __hisinf() result check * cuda : avoid warp_reduce for smax * cuda : use int instead of int64_t Noticeably improves performance (thanks to Johannes) * cuda : make loops use the same loop values Thanks Johannes again for the tip * cuda : unroll some of the loops * cuda : avoid __hisinf branches * cuda : use half2 in softmax * cuda : switch to 1 warp for bs > 16 * cuda : speed-up reduce part of the kernel * cuda : unroll Q*K^T loop * cuda : fix -INF block check * cuda : simplify softmax * cuda : fix matrix names * cuda : minor * llama : adapt to F16 KQ_pos * llama : adapt new models to F16 KQ_mask * ggml : fix F16 store (ARM NEON) * llama : fix type of KQ_mask and KQ_pos * ggml : fix CPU soft_max * tests : add hs=256 * cuda : fix build * metal : improve perf via smaller int registers * cuda : adapt soft_max to F16 mask and pos * CUDA: faster FlashAttention, kernel for bs == 1 * 16 cols for Phi-2 * no vec for hs, no hs==256 ncols==32 for Volta * adjust kernel selection logic * 4 warps, 256 stride for all D * no ncols == 64 * Multiple parallel blocks for batch size 1 * fix compile warnings * fix excessive KQ_b loads * fix cmake build * fix KV cache padding, NaN from INFINITY (#6438) * llama : flash_attn cparam + fix defrag * server: support flash_attn param * server: bench: enable flash_attn param * CUDA: refactor host code, dyn. par. blocks * fix flash_attn_vec_f16 race condition * flush softmax exp below threshold to 0 * store temp KQ in registers * Calculate KQ as FP32 if KQV has GGML_PREC_F32 * Add __hgt2_mask implementation for CUDA 11 * fix KQ FP32 precision fpr parallel_blocks > 1 * llama-bench : add -fa,--flash-attn arg * metal : add BS=1 kernel for flash attention (#6508) * metal : add BS=1 kernel for flash attention (wip) * metal : support more than 1 warps * metal : opts * metal : opt * metal : switch to parallel reduce * metal : reduce registers * metal : simplify * metal : initial FA vec kernel * metal : use F32 attention accumulators * batched-bench : add fattn arg * llama : simplify llama_build_kv_store ggml-ci * llama : adapt build_olmo to changes * ggml : fix arm fp16 store on windows * metal : clean-up * metal : clean-up kernel code * metal : minor * tests : remove benchmarks ggml-ci * ggml : fix avx512 const correctness ggml-ci * ggml : fix soft_max with bias on CPU ggml-ci * common : print --flash-attn in help * ggml : fix num dimensions in ggml_flash_attn_ext * llama : force disable flash attention for incompatible models * ggml : ggml_soft_max support F16/F32 mask/pos ggml-ci * cuda : uint -> uint32_t * cuda : "constexpr dim3" -> "const dim3" ggml-ci * cuda : try to fix __hgt2_mask ggml-ci * ggml : add TODO's for F16/F32 mask/pos support in other backends * llama : replace bool need_kq_pos with use_alibi * llama : prep ALiBi support for BERT models ggml-ci * llama : fix n_batch requirements ggml-ci * cont * server : add help for --flash-attn arg * llama : disable FA for AMD * tests : remove TMP_ATTN_BENCH ggml-ci * llama : support save/load state with FA enabled ggml-ci * ci : add CUDA save-load-state tests ggml-ci * llama : llama_kv_cache_clear zeroes data + fix save-load seq ggml-ci * llama : fix copy-paste errors, add TODO * llama : disallow incompatible states * llama : update llama_state_get_size after v_trans field * metal : remove tmp log * llama : add static reminder for llama_state_get_size * metal : fix max nsg ggml-ci * ci : fix arg order ggml-ci --------- Co-authored-by: Johannes Gäßler <johannesg@5d6.de> Co-authored-by: Pierrick HYMBERT <pierrick.hymbert@gmail.com>
2024-04-30 11:16:08 +02:00
if (src0->type == GGML_TYPE_F32 && (src1 == nullptr || src1->type == GGML_TYPE_F32) && dst->type == GGML_TYPE_F32) {
return ctx->device->pipeline_soft_max_f32;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
if (src0->type == GGML_TYPE_F32 && src1->type == GGML_TYPE_F16 && dst->type == GGML_TYPE_F32) {
return ctx->device->pipeline_soft_max_f32_f16;
}
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
return nullptr;
case GGML_OP_ROPE:
{
const int mode = ((const int32_t *) dst->op_params)[2];
const bool is_neox = mode & 2;
if (is_neox) {
if (src0->type == GGML_TYPE_F32 && dst->type == GGML_TYPE_F32) {
return ctx->device->pipeline_rope_neox_f32;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
if (src0->type == GGML_TYPE_F16 && dst->type == GGML_TYPE_F16) {
return ctx->device->pipeline_rope_neox_f16;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
} else {
if (src0->type == GGML_TYPE_F32 && dst->type == GGML_TYPE_F32) {
return ctx->device->pipeline_rope_norm_f32;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
if (src0->type == GGML_TYPE_F16 && dst->type == GGML_TYPE_F16) {
return ctx->device->pipeline_rope_norm_f16;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
}
return nullptr;
}
case GGML_OP_ARGSORT:
if (src0->type == GGML_TYPE_F32 && dst->type == GGML_TYPE_I32) {
return ctx->device->pipeline_argsort_f32;
}
return nullptr;
case GGML_OP_SUM_ROWS:
if (src0->type == GGML_TYPE_F32 && dst->type == GGML_TYPE_F32) {
return ctx->device->pipeline_sum_rows_f32;
}
return nullptr;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
default:
return nullptr;
}
GGML_UNUSED(src2);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
static ggml_vk_func_t ggml_vk_op_get_func(ggml_op op) {
switch(op) {
case GGML_OP_REPEAT:
return ggml_vk_op_repeat;
default:
return nullptr;
}
}
static bool ggml_vk_op_supports_incontiguous(ggml_op op) {
switch (op) {
case GGML_OP_CPY:
case GGML_OP_GET_ROWS:
case GGML_OP_ADD:
case GGML_OP_MUL:
case GGML_OP_DIV:
case GGML_OP_SCALE:
case GGML_OP_SQR:
case GGML_OP_CLAMP:
return true;
default:
return false;
}
}
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
template<typename PC>
static void ggml_vk_op_f32(ggml_backend_vk_context * ctx, vk_context * subctx, const ggml_tensor * src0, const ggml_tensor * src1, const ggml_tensor * src2, ggml_tensor * dst, ggml_op op, const PC&& pc) {
VK_LOG_DEBUG("ggml_vk_op_f32((" << src0 << ", name=" << src0->name << ", type=" << src0->type << ", ne0=" << src0->ne[0] << ", ne1=" << src0->ne[1] << ", ne2=" << src0->ne[2] << ", ne3=" << src0->ne[3] << ", nb0=" << src0->nb[0] << ", nb1=" << src0->nb[1] << ", nb2=" << src0->nb[2] << ", nb3=" << src0->nb[3];
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
if (src1 != nullptr) {
std::cerr << "), (" << src1 << ", name=" << src1->name << ", type=" << src1->type << ", ne0=" << src1->ne[0] << ", ne1=" << src1->ne[1] << ", ne2=" << src1->ne[2] << ", ne3=" << src1->ne[3] << ", nb0=" << src1->nb[0] << ", nb1=" << src1->nb[1] << ", nb2=" << src1->nb[2] << ", nb3=" << src1->nb[3];
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
if (src2 != nullptr) {
std::cerr << "), (" << src2 << ", name=" << src2->name << ", type=" << src2->type << ", ne0=" << src2->ne[0] << ", ne1=" << src2->ne[1] << ", ne2=" << src2->ne[2] << ", ne3=" << src2->ne[3] << ", nb0=" << src2->nb[0] << ", nb1=" << src2->nb[1] << ", nb2=" << src2->nb[2] << ", nb3=" << src2->nb[3];
}
std::cerr << "), (" << dst << ", name=" << dst->name << ", type=" << dst->type << ", ne0=" << dst->ne[0] << ", ne1=" << dst->ne[1] << ", ne2=" << dst->ne[2] << ", ne3=" << dst->ne[3] << ", nb0=" << dst->nb[0] << ", nb1=" << dst->nb[1] << ", nb2=" << dst->nb[2] << ", nb3=" << dst->nb[3] << "), " << ggml_op_name(op) << ")");
GGML_ASSERT(op == GGML_OP_GET_ROWS || (!ggml_is_quantized(src0->type) && (src1 == nullptr || !ggml_is_quantized(src1->type)))); // NOLINT
GGML_ASSERT(ggml_vk_op_supports_incontiguous(op) || ggml_vk_dim01_contiguous(src0)); // NOLINT
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
GGML_ASSERT(dst->extra != nullptr);
const uint64_t ne00 = src0->ne[0];
const uint64_t ne01 = src0->ne[1];
const uint64_t ne02 = src0->ne[2];
const uint64_t ne03 = src0->ne[3];
const uint64_t ne0 = ne00 * ne01;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
const bool use_src1 = src1 != nullptr;
const uint64_t ne10 = use_src1 ? src1->ne[0] : 0;
const uint64_t ne11 = use_src1 ? src1->ne[1] : 0;
const uint64_t ne12 = use_src1 ? src1->ne[2] : 0;
const uint64_t ne13 = use_src1 ? src1->ne[3] : 0;
const uint64_t ne1 = ne10 * ne11;
// const uint64_t nb10 = use_src1 ? src1->nb[0] : 0;
const bool use_src2 = src2 != nullptr;
const uint64_t ne20 = use_src2 ? src2->ne[0] : 0;
const uint64_t ne21 = use_src2 ? src2->ne[1] : 0;
const uint64_t ne22 = use_src2 ? src2->ne[2] : 0;
const uint64_t ne23 = use_src2 ? src2->ne[3] : 0;
const uint64_t ne2 = ne20 * ne21;
const uint64_t ned0 = dst->ne[0];
const uint64_t ned1 = dst->ne[1];
const uint64_t ned2 = dst->ne[2];
const uint64_t ned3 = dst->ne[3];
const uint64_t ned = ned0 * ned1;
vk_pipeline pipeline = ggml_vk_op_get_pipeline(ctx, src0, src1, src2, dst, op);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
ggml_vk_func_t op_func;
if (pipeline == nullptr) {
op_func = ggml_vk_op_get_func(op);
if (op_func == nullptr) {
std::cerr << "ggml_vulkan: Error: Missing op: " << ggml_op_name(op) << " for " << ggml_type_name(src0->type);
if (src1 != nullptr) {
std::cerr << " and " << ggml_type_name(src1->type);
}
std::cerr << " to " << ggml_type_name(dst->type) << std::endl;
GGML_ASSERT(false);
}
op_func(ctx, subctx, src0, src1, dst);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
return;
}
const bool op_supports_incontiguous = ggml_vk_op_supports_incontiguous(op);
ggml_tensor_extra_gpu * extra = (ggml_tensor_extra_gpu *) dst->extra;
ggml_tensor_extra_gpu * extra_src0 = (ggml_tensor_extra_gpu *) src0->extra;
ggml_tensor_extra_gpu * extra_src1 = use_src1 ? (ggml_tensor_extra_gpu *) src1->extra : nullptr;
ggml_tensor_extra_gpu * extra_src2 = use_src2 ? (ggml_tensor_extra_gpu *) src2->extra : nullptr;
vk_buffer d_X = nullptr;
size_t x_buf_offset = 0;
vk_buffer d_Y = nullptr;
size_t y_buf_offset = 0;
vk_buffer d_Z = nullptr;
size_t z_buf_offset = 0;
bool src0_uma = false;
bool src1_uma = false;
bool src2_uma = false;
if (ctx->device->uma) {
ggml_vk_host_get(ctx->device, src0->data, d_X, x_buf_offset);
src0_uma = d_X != nullptr;
if (use_src1) {
ggml_vk_host_get(ctx->device, src1->data, d_Y, y_buf_offset);
src1_uma = d_Y != nullptr;
}
if (use_src2) {
ggml_vk_host_get(ctx->device, src2->data, d_Z, z_buf_offset);
src2_uma = d_Z != nullptr;
}
}
uint64_t x_sz = ggml_type_size(src0->type)/ggml_blck_size(src0->type) * ne0;
uint64_t y_sz = use_src1 ? ggml_type_size(src1->type) * ne1 : 0;
uint64_t z_sz = use_src2 ? ggml_type_size(src2->type) * ne2 : 0;
uint64_t d_sz = ggml_type_size(dst->type) * ned;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
vk_buffer d_D = extra->buffer_gpu.lock();
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
// Workaround for tiny tensor inputs on ROPE
if (use_src1 && y_sz > d_D->size) {
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
y_sz = VK_WHOLE_SIZE;
}
GGML_ASSERT(d_D != nullptr);
uint64_t d_buf_offset = ((extra->offset + dst->view_offs) / ctx->device->properties.limits.minStorageBufferOffsetAlignment) * ctx->device->properties.limits.minStorageBufferOffsetAlignment;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
GGML_ASSERT(d_buf_offset == extra->offset || op == GGML_OP_CPY); // NOLINT
if(!src0_uma) {
d_X = extra_src0->buffer_gpu.lock();
x_buf_offset = extra_src0->offset + src0->view_offs;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
GGML_ASSERT(d_X != nullptr);
}
if (use_src1 && !src1_uma) {
d_Y = extra_src1->buffer_gpu.lock();
y_buf_offset = extra_src1->offset + src1->view_offs;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
GGML_ASSERT(d_Y != nullptr);
}
if (use_src2 && !src2_uma) {
d_Z = extra_src2->buffer_gpu.lock();
z_buf_offset = extra_src2->offset + src2->view_offs;
GGML_ASSERT(d_Z != nullptr);
}
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
if (op_supports_incontiguous) {
x_sz = ggml_nbytes(src0);
y_sz = use_src1 ? ggml_nbytes(src1) : 0;
z_sz = use_src2 ? ggml_nbytes(src2) : 0;
d_sz = ggml_nbytes(dst);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
if (x_buf_offset + x_sz >= d_X->size) {
x_sz = VK_WHOLE_SIZE;
}
if (use_src1 && y_buf_offset + y_sz >= d_Y->size) {
y_sz = VK_WHOLE_SIZE;
}
if (use_src2 && z_buf_offset + z_sz >= d_Z->size) {
z_sz = VK_WHOLE_SIZE;
}
if (d_buf_offset + d_sz >= d_D->size) {
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
d_sz = VK_WHOLE_SIZE;
}
}
std::array<uint32_t, 3> elements;
// Single call if dimension 2 is contiguous
if (op_supports_incontiguous || (ggml_is_contiguous(src0) && (src1 == nullptr || ggml_is_contiguous(src1)))) {
ggml_pipeline_allocate_descriptor_sets(ctx->device, pipeline, 1);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
switch (dst->op) {
case GGML_OP_NORM:
case GGML_OP_RMS_NORM:
case GGML_OP_SOFT_MAX:
case GGML_OP_SUM_ROWS:
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
elements = { (uint32_t)ggml_nrows(src0), 1, 1 };
break;
case GGML_OP_DIAG_MASK_INF:
case GGML_OP_ROPE:
elements = { (uint32_t)ggml_nrows(src0), (uint32_t)ne00, 1 };
break;
case GGML_OP_GET_ROWS:
elements = { (uint32_t)ne00, (uint32_t)ne10, (uint32_t)(ne11 * ne12) };
break;
case GGML_OP_ARGSORT:
elements = { (uint32_t)ne00, (uint32_t)ggml_nrows(src0), 1 };
break;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
default:
elements = { (uint32_t)ggml_nelements(src0), 1, 1 };
break;
}
if (!op_supports_incontiguous) {
if (x_sz != VK_WHOLE_SIZE) {
x_sz *= ne02 * ne03;
}
if (use_src1 && y_sz != VK_WHOLE_SIZE) {
y_sz *= ne12 * ne13;
}
if (use_src2 && z_sz != VK_WHOLE_SIZE) {
z_sz *= ne22 * ne23;
}
if (d_sz != VK_WHOLE_SIZE) {
d_sz *= ned2 * ned3;
}
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
if (op == GGML_OP_SOFT_MAX) {
// Empty src1 is possible in soft_max, but the shader needs a buffer
vk_subbuffer subbuf_y;
if (use_src1) {
subbuf_y = { d_Y, y_buf_offset, y_sz };
} else {
subbuf_y = { d_X, 0, d_X->size };
}
ggml_vk_sync_buffers(subctx);
ggml_vk_dispatch_pipeline(ctx, subctx, pipeline, { { d_X, x_buf_offset, x_sz }, subbuf_y, { d_D, d_buf_offset, d_sz } }, sizeof(PC), &pc, elements);
} else if (op == GGML_OP_ROPE) {
// Empty src2 is possible in rope, but the shader needs a buffer
vk_subbuffer subbuf_z;
if (use_src2) {
subbuf_z = { d_Z, z_buf_offset, z_sz };
} else {
subbuf_z = { d_X, 0, d_X->size };
}
ggml_vk_sync_buffers(subctx);
ggml_vk_dispatch_pipeline(ctx, subctx, pipeline, { { d_X, x_buf_offset, x_sz }, { d_Y, y_buf_offset, y_sz }, subbuf_z, { d_D, d_buf_offset, d_sz } }, sizeof(PC), &pc, elements);
} else if (use_src2) {
ggml_vk_sync_buffers(subctx);
ggml_vk_dispatch_pipeline(ctx, subctx, pipeline, { { d_X, x_buf_offset, x_sz }, { d_Y, y_buf_offset, y_sz }, { d_Z, z_buf_offset, z_sz }, { d_D, d_buf_offset, d_sz } }, sizeof(PC), &pc, elements);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
} else if (use_src1) {
ggml_vk_sync_buffers(subctx);
ggml_vk_dispatch_pipeline(ctx, subctx, pipeline, { { d_X, x_buf_offset, x_sz }, { d_Y, y_buf_offset, y_sz }, { d_D, d_buf_offset, d_sz } }, sizeof(PC), &pc, elements);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
} else {
ggml_vk_sync_buffers(subctx);
ggml_vk_dispatch_pipeline(ctx, subctx, pipeline, { { d_X, x_buf_offset, x_sz }, { d_D, d_buf_offset, d_sz } }, sizeof(PC), &pc, elements);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
} else {
GGML_ASSERT(op != GGML_OP_SOFT_MAX);
GGML_ASSERT(op != GGML_OP_ARGSORT);
GGML_ASSERT(!use_src2);
ggml_pipeline_allocate_descriptor_sets(ctx->device, pipeline, ne02 * ne03);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
switch (dst->op) {
case GGML_OP_NORM:
case GGML_OP_RMS_NORM:
elements = { (uint32_t)ne01, 1, 1 };
break;
case GGML_OP_DIAG_MASK_INF:
case GGML_OP_ROPE:
elements = { (uint32_t)ne01, (uint32_t)ne00, 1 };
break;
case GGML_OP_GET_ROWS:
elements = { (uint32_t)ne00, (uint32_t)ne10, (uint32_t)(ne11 * ne12) };
break;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
default:
elements = { (uint32_t)ne0, 1, 1 };
break;
}
for (uint64_t i03 = 0; i03 < ne03; i03++) {
for (uint64_t i02 = 0; i02 < ne02; i02++) {
const uint32_t it_idx0 = (i03 * ne02 + i02);
const uint32_t it_idx1 = use_src1 ? ((i03 % ne13) * ne12 + (i02 % ne12)) : 0;
const uint32_t x_offset = x_sz * it_idx0;
const uint32_t y_offset = y_sz * it_idx1;
const uint32_t d_offset = d_sz * it_idx0;
if (use_src1) {
ggml_vk_sync_buffers(subctx);
ggml_vk_dispatch_pipeline(ctx, subctx, pipeline, { { d_X, x_buf_offset + x_offset, x_sz }, { d_Y, y_buf_offset + y_offset, y_sz }, { d_D, d_buf_offset + d_offset, d_sz } }, sizeof(PC), &pc, elements);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
} else {
ggml_vk_sync_buffers(subctx);
ggml_vk_dispatch_pipeline(ctx, subctx, pipeline, { { d_X, x_buf_offset + x_offset, x_sz }, { d_D, d_buf_offset + d_offset, d_sz } }, sizeof(PC), &pc, elements);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
}
}
}
}
static void ggml_vk_repeat(ggml_backend_vk_context * ctx, vk_context * subctx, const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst) {
ggml_vk_op_f32<vk_op_push_constants>(ctx, subctx, src0, src1, nullptr, dst, GGML_OP_REPEAT, { (uint32_t)ggml_nelements(src0), (uint32_t)ggml_nelements(src1), 0.0f, 0.0f });
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
static void ggml_vk_get_rows(ggml_backend_vk_context * ctx, vk_context * subctx, const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst) {
const uint32_t src0_type_size = ggml_type_size(src0->type);
const uint32_t src1_type_size = ggml_type_size(src1->type);
const uint32_t dst_type_size = ggml_type_size(dst->type);
ggml_vk_op_f32<vk_op_binary_push_constants>(ctx, subctx, src0, src1, nullptr, dst, GGML_OP_GET_ROWS, {
(uint32_t)ggml_nelements(src0),
(uint32_t)src0->ne[0], (uint32_t)src0->ne[1], (uint32_t)src0->ne[2],(uint32_t)src0->ne[3], (uint32_t)src0->nb[0] / src0_type_size, (uint32_t)src0->nb[1] / src0_type_size, (uint32_t)src0->nb[2] / src0_type_size, (uint32_t)src0->nb[3] / src0_type_size,
(uint32_t)src1->ne[0], (uint32_t)src1->ne[1], (uint32_t)src1->ne[2],(uint32_t)src1->ne[3], (uint32_t)src1->nb[0] / src1_type_size, (uint32_t)src1->nb[1] / src1_type_size, (uint32_t)src1->nb[2] / src1_type_size, (uint32_t)src1->nb[3] / src1_type_size,
(uint32_t) dst->ne[0], (uint32_t) dst->ne[1], (uint32_t) dst->ne[2],(uint32_t) dst->ne[3], (uint32_t) dst->nb[0] / dst_type_size, (uint32_t) dst->nb[1] / dst_type_size, (uint32_t) dst->nb[2] / dst_type_size, (uint32_t) dst->nb[3] / dst_type_size,
0,
0.0f, 0.0f,
});
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
static void ggml_vk_add(ggml_backend_vk_context * ctx, vk_context * subctx, const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst) {
const uint32_t src0_type_size = ggml_type_size(src0->type);
const uint32_t src1_type_size = ggml_type_size(src1->type);
const uint32_t dst_type_size = ggml_type_size(dst->type);
ggml_vk_op_f32<vk_op_binary_push_constants>(ctx, subctx, src0, src1, nullptr, dst, GGML_OP_ADD, {
(uint32_t)ggml_nelements(src0),
(uint32_t)src0->ne[0], (uint32_t)src0->ne[1], (uint32_t)src0->ne[2],(uint32_t)src0->ne[3], (uint32_t)src0->nb[0] / src0_type_size, (uint32_t)src0->nb[1] / src0_type_size, (uint32_t)src0->nb[2] / src0_type_size, (uint32_t)src0->nb[3] / src0_type_size,
(uint32_t)src1->ne[0], (uint32_t)src1->ne[1], (uint32_t)src1->ne[2],(uint32_t)src1->ne[3], (uint32_t)src1->nb[0] / src1_type_size, (uint32_t)src1->nb[1] / src1_type_size, (uint32_t)src1->nb[2] / src1_type_size, (uint32_t)src1->nb[3] / src1_type_size,
(uint32_t) dst->ne[0], (uint32_t) dst->ne[1], (uint32_t) dst->ne[2],(uint32_t) dst->ne[3], (uint32_t) dst->nb[0] / dst_type_size, (uint32_t) dst->nb[1] / dst_type_size, (uint32_t) dst->nb[2] / dst_type_size, (uint32_t) dst->nb[3] / dst_type_size,
0,
0.0f, 0.0f,
});
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
static void ggml_vk_mul(ggml_backend_vk_context * ctx, vk_context * subctx, const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst) {
const uint32_t src0_type_size = ggml_type_size(src0->type);
const uint32_t src1_type_size = ggml_type_size(src1->type);
const uint32_t dst_type_size = ggml_type_size(dst->type);
ggml_vk_op_f32<vk_op_binary_push_constants>(ctx, subctx, src0, src1, nullptr, dst, GGML_OP_MUL, {
(uint32_t)ggml_nelements(src0),
(uint32_t)src0->ne[0], (uint32_t)src0->ne[1], (uint32_t)src0->ne[2],(uint32_t)src0->ne[3], (uint32_t)src0->nb[0] / src0_type_size, (uint32_t)src0->nb[1] / src0_type_size, (uint32_t)src0->nb[2] / src0_type_size, (uint32_t)src0->nb[3] / src0_type_size,
(uint32_t)src1->ne[0], (uint32_t)src1->ne[1], (uint32_t)src1->ne[2],(uint32_t)src1->ne[3], (uint32_t)src1->nb[0] / src1_type_size, (uint32_t)src1->nb[1] / src1_type_size, (uint32_t)src1->nb[2] / src1_type_size, (uint32_t)src1->nb[3] / src1_type_size,
(uint32_t) dst->ne[0], (uint32_t) dst->ne[1], (uint32_t) dst->ne[2],(uint32_t) dst->ne[3], (uint32_t) dst->nb[0] / dst_type_size, (uint32_t) dst->nb[1] / dst_type_size, (uint32_t) dst->nb[2] / dst_type_size, (uint32_t) dst->nb[3] / dst_type_size,
0,
0.0f, 0.0f,
});
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
static void ggml_vk_div(ggml_backend_vk_context * ctx, vk_context * subctx, const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst) {
const uint32_t src0_type_size = ggml_type_size(src0->type);
const uint32_t src1_type_size = ggml_type_size(src1->type);
const uint32_t dst_type_size = ggml_type_size(dst->type);
ggml_vk_op_f32<vk_op_binary_push_constants>(ctx, subctx, src0, src1, nullptr, dst, GGML_OP_DIV, {
(uint32_t)ggml_nelements(src0),
(uint32_t)src0->ne[0], (uint32_t)src0->ne[1], (uint32_t)src0->ne[2],(uint32_t)src0->ne[3], (uint32_t)src0->nb[0] / src0_type_size, (uint32_t)src0->nb[1] / src0_type_size, (uint32_t)src0->nb[2] / src0_type_size, (uint32_t)src0->nb[3] / src0_type_size,
(uint32_t)src1->ne[0], (uint32_t)src1->ne[1], (uint32_t)src1->ne[2],(uint32_t)src1->ne[3], (uint32_t)src1->nb[0] / src1_type_size, (uint32_t)src1->nb[1] / src1_type_size, (uint32_t)src1->nb[2] / src1_type_size, (uint32_t)src1->nb[3] / src1_type_size,
(uint32_t) dst->ne[0], (uint32_t) dst->ne[1], (uint32_t) dst->ne[2],(uint32_t) dst->ne[3], (uint32_t) dst->nb[0] / dst_type_size, (uint32_t) dst->nb[1] / dst_type_size, (uint32_t) dst->nb[2] / dst_type_size, (uint32_t) dst->nb[3] / dst_type_size,
0,
0.0f, 0.0f,
});
}
static void ggml_vk_scale(ggml_backend_vk_context * ctx, vk_context * subctx, const ggml_tensor * src0, ggml_tensor * dst) {
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
float * op_params = (float *)dst->op_params;
const uint32_t src0_type_size = ggml_type_size(src0->type);
const uint32_t dst_type_size = ggml_type_size(dst->type);
ggml_vk_op_f32<vk_op_unary_push_constants>(ctx, subctx, src0, nullptr, nullptr, dst, GGML_OP_SCALE, {
(uint32_t)ggml_nelements(src0),
(uint32_t)src0->ne[0], (uint32_t)src0->ne[1], (uint32_t)src0->ne[2], (uint32_t)src0->ne[3], (uint32_t)src0->nb[0] / src0_type_size, (uint32_t)src0->nb[1] / src0_type_size, (uint32_t)src0->nb[2] / src0_type_size, (uint32_t)src0->nb[3] / src0_type_size,
(uint32_t) dst->ne[0], (uint32_t) dst->ne[1], (uint32_t) dst->ne[2], (uint32_t) dst->ne[3], (uint32_t) dst->nb[0] / dst_type_size, (uint32_t) dst->nb[1] / dst_type_size, (uint32_t) dst->nb[2] / dst_type_size, (uint32_t) dst->nb[3] / dst_type_size,
0,
op_params[0], 0.0f
});
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
static void ggml_vk_sqr(ggml_backend_vk_context * ctx, vk_context * subctx, const ggml_tensor * src0, ggml_tensor * dst) {
const uint32_t src0_type_size = ggml_type_size(src0->type);
const uint32_t dst_type_size = ggml_type_size(dst->type);
ggml_vk_op_f32<vk_op_unary_push_constants>(ctx, subctx, src0, nullptr, nullptr, dst, GGML_OP_SQR, {
(uint32_t)ggml_nelements(src0),
(uint32_t)src0->ne[0], (uint32_t)src0->ne[1], (uint32_t)src0->ne[2], (uint32_t)src0->ne[3], (uint32_t)src0->nb[0] / src0_type_size, (uint32_t)src0->nb[1] / src0_type_size, (uint32_t)src0->nb[2] / src0_type_size, (uint32_t)src0->nb[3] / src0_type_size,
(uint32_t) dst->ne[0], (uint32_t) dst->ne[1], (uint32_t) dst->ne[2], (uint32_t) dst->ne[3], (uint32_t) dst->nb[0] / dst_type_size, (uint32_t) dst->nb[1] / dst_type_size, (uint32_t) dst->nb[2] / dst_type_size, (uint32_t) dst->nb[3] / dst_type_size,
0,
0.0f, 0.0f,
});
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
static void ggml_vk_clamp(ggml_backend_vk_context * ctx, vk_context * subctx, const ggml_tensor * src0, ggml_tensor * dst) {
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
float * op_params = (float *)dst->op_params;
const uint32_t src0_type_size = ggml_type_size(src0->type);
const uint32_t dst_type_size = ggml_type_size(dst->type);
ggml_vk_op_f32<vk_op_unary_push_constants>(ctx, subctx, src0, nullptr, nullptr, dst, GGML_OP_CLAMP, {
(uint32_t)ggml_nelements(src0),
(uint32_t)src0->ne[0], (uint32_t)src0->ne[1], (uint32_t)src0->ne[2], (uint32_t)src0->ne[3], (uint32_t)src0->nb[0] / src0_type_size, (uint32_t)src0->nb[1] / src0_type_size, (uint32_t)src0->nb[2] / src0_type_size, (uint32_t)src0->nb[3] / src0_type_size,
(uint32_t) dst->ne[0], (uint32_t) dst->ne[1], (uint32_t) dst->ne[2], (uint32_t) dst->ne[3], (uint32_t) dst->nb[0] / dst_type_size, (uint32_t) dst->nb[1] / dst_type_size, (uint32_t) dst->nb[2] / dst_type_size, (uint32_t) dst->nb[3] / dst_type_size,
0,
op_params[0], op_params[1],
});
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
static void ggml_vk_cpy(ggml_backend_vk_context * ctx, vk_context * subctx, const ggml_tensor * src0, ggml_tensor * dst) {
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
ggml_tensor_extra_gpu * extra = (ggml_tensor_extra_gpu *) dst->extra;
const uint32_t src0_type_size = ggml_type_size(src0->type);
const uint32_t dst_type_size = ggml_type_size(dst->type);
const uint32_t d_offset = ((extra->offset + dst->view_offs) % ctx->device->properties.limits.minStorageBufferOffsetAlignment) / dst_type_size;
ggml_vk_op_f32<vk_op_unary_push_constants>(ctx, subctx, src0, nullptr, nullptr, dst, GGML_OP_CPY, {
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
(uint32_t)ggml_nelements(src0),
(uint32_t)src0->ne[0], (uint32_t)src0->ne[1], (uint32_t)src0->ne[2], (uint32_t)src0->ne[3], (uint32_t)src0->nb[0] / src0_type_size, (uint32_t)src0->nb[1] / src0_type_size, (uint32_t)src0->nb[2] / src0_type_size, (uint32_t)src0->nb[3] / src0_type_size,
(uint32_t) dst->ne[0], (uint32_t) dst->ne[1], (uint32_t) dst->ne[2], (uint32_t) dst->ne[3], (uint32_t) dst->nb[0] / dst_type_size, (uint32_t) dst->nb[1] / dst_type_size, (uint32_t) dst->nb[2] / dst_type_size, (uint32_t) dst->nb[3] / dst_type_size,
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
d_offset,
0.0f, 0.0f,
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
});
}
static void ggml_vk_norm(ggml_backend_vk_context * ctx, vk_context * subctx, const ggml_tensor * src0, ggml_tensor * dst) {
float * op_params = (float *)dst->op_params;
ggml_vk_op_f32<vk_op_push_constants>(ctx, subctx, src0, nullptr, nullptr, dst, GGML_OP_NORM, { (uint32_t)src0->ne[0], (uint32_t)src0->ne[1], op_params[0], 0.0f });
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
static void ggml_vk_rms_norm(ggml_backend_vk_context * ctx, vk_context * subctx, const ggml_tensor * src0, ggml_tensor * dst) {
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
float * op_params = (float *)dst->op_params;
ggml_vk_op_f32<vk_op_push_constants>(ctx, subctx, src0, nullptr, nullptr, dst, GGML_OP_RMS_NORM, { (uint32_t)src0->ne[0], (uint32_t)src0->ne[1], op_params[0], 0.0f });
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
static void ggml_vk_unary(ggml_backend_vk_context * ctx, vk_context * subctx, const ggml_tensor * src0, ggml_tensor * dst) {
ggml_vk_op_f32<vk_op_push_constants>(ctx, subctx, src0, nullptr, nullptr, dst, GGML_OP_UNARY, { (uint32_t)ggml_nelements(src0), 0, 0.0f, 0.0f });
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
static void ggml_vk_diag_mask_inf(ggml_backend_vk_context * ctx, vk_context * subctx, const ggml_tensor * src0, ggml_tensor * dst) {
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
int32_t * op_params = (int32_t *)dst->op_params;
ggml_vk_op_f32<vk_op_diag_mask_push_constants>(ctx, subctx, src0, nullptr, nullptr, dst, GGML_OP_DIAG_MASK_INF, { (uint32_t)src0->ne[0], (uint32_t)src0->ne[1], op_params[0] });
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
static void ggml_vk_soft_max(ggml_backend_vk_context * ctx, vk_context * subctx, const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst) {
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
float * op_params = (float *)dst->op_params;
float scale = op_params[0];
float max_bias = op_params[1];
const uint32_t ncols = (uint32_t)src0->ne[0];
const uint32_t nrows_x = (uint32_t)ggml_nrows(src0);
const uint32_t nrows_y = (uint32_t)src0->ne[1];
const uint32_t n_head_kv = nrows_x/nrows_y;
const uint32_t n_head_log2 = 1u << (uint32_t) floorf(log2f((float) n_head_kv));
const float m0 = powf(2.0f, -(max_bias ) / n_head_log2);
const float m1 = powf(2.0f, -(max_bias / 2.0f) / n_head_log2);
ggml_vk_op_f32<vk_op_soft_max_push_constants>(ctx, subctx, src0, src1, nullptr, dst, GGML_OP_SOFT_MAX, {
ncols,
src1 != nullptr ? nrows_y : (uint32_t)0,
scale, max_bias,
m0, m1,
n_head_log2,
});
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
static void ggml_vk_rope(ggml_backend_vk_context * ctx, vk_context * subctx, const ggml_tensor * src0, const ggml_tensor * src1, const ggml_tensor * src2, ggml_tensor * dst) {
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
const int n_dims = ((int32_t *) dst->op_params)[1];
// const int mode = ((int32_t *) dst->op_params)[2];
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
// const int n_ctx = ((int32_t *) dst->op_params)[3];
const int n_ctx_orig = ((int32_t *) dst->op_params)[4];
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
const float freq_base = ((float *) dst->op_params)[5];
const float freq_scale = ((float *) dst->op_params)[6];
const float ext_factor = ((float *) dst->op_params)[7];
const float attn_factor = ((float *) dst->op_params)[8];
const float beta_fast = ((float *) dst->op_params)[9];
const float beta_slow = ((float *) dst->op_params)[10];
float corr_dims[2];
ggml_rope_yarn_corr_dims(n_dims, n_ctx_orig, freq_base, beta_fast, beta_slow, corr_dims);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
const float theta_scale = powf(freq_base, -2.0f/n_dims);
ggml_vk_op_f32<vk_op_rope_push_constants>(ctx, subctx, src0, src1, src2, dst, GGML_OP_ROPE, {
(uint32_t)src0->ne[0], (uint32_t)n_dims, freq_scale, (uint32_t)src0->ne[1],
freq_base, ext_factor, attn_factor, {corr_dims[0], corr_dims[1]}, theta_scale,
src2 != nullptr,
});
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
static void ggml_vk_argsort(ggml_backend_vk_context * ctx, vk_context * subctx, const ggml_tensor * src0, ggml_tensor * dst) {
int32_t * op_params = (int32_t *)dst->op_params;
uint32_t ncols = src0->ne[0];
uint32_t ncols_pad = 1;
while (ncols_pad < ncols) {
ncols_pad *= 2;
}
GGML_ASSERT(ncols_pad <= 1024);
ggml_vk_op_f32<vk_op_argsort_push_constants>(ctx, subctx, src0, nullptr, nullptr, dst, GGML_OP_ARGSORT, {
ncols,
ncols_pad,
op_params[0],
});
}
static void ggml_vk_sum_rows(ggml_backend_vk_context * ctx, vk_context * subctx, const ggml_tensor * src0, ggml_tensor * dst) {
ggml_vk_op_f32<vk_op_push_constants>(ctx, subctx, src0, nullptr, nullptr, dst, GGML_OP_SUM_ROWS, { (uint32_t)src0->ne[0], 0, 0.0f, 0.0f });
}
#ifdef GGML_VULKAN_RUN_TESTS
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
static void ggml_vk_print_matrix_area(const void * data, ggml_type type, int ne0, int ne1, int i0, int i1, int i2) {
if (type != GGML_TYPE_F32 && type != GGML_TYPE_F16) {
return;
}
i0 = std::max(i0, 5);
i1 = std::max(i1, 5);
i2 = std::max(i2, 0);
fprintf(stderr, " ");
for (int idx1 = i1 - 5; idx1 < i1 + 5; idx1++) {
fprintf(stderr, "%7d ", idx1);
}
fprintf(stderr, "\n");
for (int idx0 = i0 - 5; idx0 < i0 + 5; idx0++) {
fprintf(stderr, "%7d: ", idx0);
for (int idx1 = i1 - 5; idx1 < i1 + 5; idx1++) {
if (idx0 >= 0 && idx0 < ne0 && idx1 >= 0 && idx1 < ne1) {
float val;
if (type == GGML_TYPE_F32) {
val = *((const float *) data + i2*ne1*ne0 + idx1*ne0 + idx0);
} else if (type == GGML_TYPE_F16) {
val = ggml_fp16_to_fp32(*((const ggml_fp16_t *) data + i2*ne1*ne0 + idx1*ne0 + idx0));
} else {
GGML_ASSERT(false);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
fprintf(stderr, "% 7.2f ", val);
} else {
fprintf(stderr, " ");
}
}
fprintf(stderr, "\n");
}
}
template <typename X_TYPE, typename Y_TYPE>
static void ggml_vk_test_matmul(ggml_backend_vk_context * ctx, size_t m, size_t n, size_t k, size_t batch, size_t num_it, int split_k, int shader_size) {
VK_LOG_DEBUG("ggml_vk_test_matmul(" << m << ", " << n << ", " << k << ", " << batch << ", " << num_it << ", " << split_k << ", " << shader_size << ")");
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
const size_t x_ne = m * k * batch;
const size_t y_ne = k * n * batch;
const size_t d_ne = m * n * batch;
vk_pipeline p;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
std::string shname;
if (shader_size == 0) {
if (std::is_same<float, X_TYPE>() && std::is_same<float, Y_TYPE>()) {
p = ctx->device->pipeline_matmul_f32->a_s;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
shname = "F32_ALIGNED_S";
} else if (std::is_same<float, X_TYPE>() && std::is_same<ggml_fp16_t, Y_TYPE>()) {
p = ctx->device->pipeline_matmul_f32_f16->a_s;
shname = "F32_F16_ALIGNED_S";
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
} else if (std::is_same<ggml_fp16_t, X_TYPE>() && std::is_same<float, Y_TYPE>()) {
p = ctx->device->pipeline_matmul_f16_f32->a_s;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
shname = "F16_F32_ALIGNED_S";
} else if (std::is_same<ggml_fp16_t, X_TYPE>() && std::is_same<ggml_fp16_t, Y_TYPE>()) {
p = ctx->device->pipeline_matmul_f16->a_s;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
shname = "F16_ALIGNED_S";
} else {
GGML_ASSERT(false);
}
} else if (shader_size == 1) {
if (std::is_same<float, X_TYPE>() && std::is_same<float, Y_TYPE>()) {
p = ctx->device->pipeline_matmul_f32->a_m;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
shname = "F32_ALIGNED_M";
} else if (std::is_same<float, X_TYPE>() && std::is_same<ggml_fp16_t, Y_TYPE>()) {
p = ctx->device->pipeline_matmul_f32_f16->a_m;
shname = "F32_F16_ALIGNED_M";
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
} else if (std::is_same<ggml_fp16_t, X_TYPE>() && std::is_same<float, Y_TYPE>()) {
p = ctx->device->pipeline_matmul_f16_f32->a_m;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
shname = "F16_F32_ALIGNED_M";
} else if (std::is_same<ggml_fp16_t, X_TYPE>() && std::is_same<ggml_fp16_t, Y_TYPE>()) {
p = ctx->device->pipeline_matmul_f16->a_m;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
shname = "F16_ALIGNED_M";
} else {
GGML_ASSERT(false);
}
} else if (shader_size == 2) {
if (std::is_same<float, X_TYPE>() && std::is_same<float, Y_TYPE>()) {
p = ctx->device->pipeline_matmul_f32->a_l;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
shname = "F32_ALIGNED_L";
} else if (std::is_same<float, X_TYPE>() && std::is_same<ggml_fp16_t, Y_TYPE>()) {
p = ctx->device->pipeline_matmul_f32_f16->a_l;
shname = "F32_F16_ALIGNED_L";
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
} else if (std::is_same<ggml_fp16_t, X_TYPE>() && std::is_same<float, Y_TYPE>()) {
p = ctx->device->pipeline_matmul_f16_f32->a_l;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
shname = "F16_F32_ALIGNED_L";
} else if (std::is_same<ggml_fp16_t, X_TYPE>() && std::is_same<ggml_fp16_t, Y_TYPE>()) {
p = ctx->device->pipeline_matmul_f16->a_l;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
shname = "F16_ALIGNED_L";
} else {
GGML_ASSERT(false);
}
} else {
GGML_ASSERT(0);
}
const size_t kpad = ggml_vk_align_size(k, p->align);
if (k != kpad) {
if (shader_size == 0) {
if (std::is_same<float, X_TYPE>() && std::is_same<float, Y_TYPE>()) {
p = ctx->device->pipeline_matmul_f32->s;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
shname = "F32_S";
} else if (std::is_same<float, X_TYPE>() && std::is_same<ggml_fp16_t, Y_TYPE>()) {
p = ctx->device->pipeline_matmul_f32_f16->s;
shname = "F32_F16_S";
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
} else if (std::is_same<ggml_fp16_t, X_TYPE>() && std::is_same<float, Y_TYPE>()) {
p = ctx->device->pipeline_matmul_f16_f32->s;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
shname = "F16_F32_S";
} else if (std::is_same<ggml_fp16_t, X_TYPE>() && std::is_same<ggml_fp16_t, Y_TYPE>()) {
p = ctx->device->pipeline_matmul_f16->s;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
shname = "F16_S";
}
} else if (shader_size == 1) {
if (std::is_same<float, X_TYPE>() && std::is_same<float, Y_TYPE>()) {
p = ctx->device->pipeline_matmul_f32->m;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
shname = "F32_M";
} else if (std::is_same<float, X_TYPE>() && std::is_same<ggml_fp16_t, Y_TYPE>()) {
p = ctx->device->pipeline_matmul_f32_f16->m;
shname = "F32_F16_M";
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
} else if (std::is_same<ggml_fp16_t, X_TYPE>() && std::is_same<float, Y_TYPE>()) {
p = ctx->device->pipeline_matmul_f16_f32->m;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
shname = "F16_F32_M";
} else if (std::is_same<ggml_fp16_t, X_TYPE>() && std::is_same<ggml_fp16_t, Y_TYPE>()) {
p = ctx->device->pipeline_matmul_f16->m;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
shname = "F16_M";
}
} else if (shader_size == 2) {
if (std::is_same<float, X_TYPE>() && std::is_same<float, Y_TYPE>()) {
p = ctx->device->pipeline_matmul_f32->l;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
shname = "F32_L";
} else if (std::is_same<float, X_TYPE>() && std::is_same<ggml_fp16_t, Y_TYPE>()) {
p = ctx->device->pipeline_matmul_f32_f16->l;
shname = "F32_F16_L";
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
} else if (std::is_same<ggml_fp16_t, X_TYPE>() && std::is_same<float, Y_TYPE>()) {
p = ctx->device->pipeline_matmul_f16_f32->l;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
shname = "F16_F32_L";
} else if (std::is_same<ggml_fp16_t, X_TYPE>() && std::is_same<ggml_fp16_t, Y_TYPE>()) {
p = ctx->device->pipeline_matmul_f16->l;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
shname = "F16_L";
}
}
}
ggml_pipeline_allocate_descriptor_sets(ctx, p, num_it);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
if (split_k > 1) {
ggml_pipeline_allocate_descriptor_sets(ctx, ctx->device->pipeline_matmul_split_k_reduce, num_it);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
if (ctx->prealloc_split_k == nullptr || ctx->prealloc_split_k->size < sizeof(float) * d_ne * split_k) {
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
// Resize buffer
if (ctx->prealloc_split_k != nullptr) {
ggml_vk_destroy_buffer(ctx->prealloc_split_k);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
ctx->prealloc_split_k = ggml_vk_create_buffer_check(ctx, sizeof(float) * d_ne * split_k, vk::MemoryPropertyFlagBits::eDeviceLocal);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
}
vk_buffer d_X = ggml_vk_create_buffer_check(ctx, sizeof(X_TYPE) * x_ne, vk::MemoryPropertyFlagBits::eDeviceLocal);
vk_buffer d_Y = ggml_vk_create_buffer_check(ctx, sizeof(Y_TYPE) * y_ne, vk::MemoryPropertyFlagBits::eDeviceLocal);
vk_buffer d_D = ggml_vk_create_buffer_check(ctx, sizeof(float) * d_ne, vk::MemoryPropertyFlagBits::eDeviceLocal);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
X_TYPE* x = (X_TYPE *) malloc(sizeof(X_TYPE) * x_ne);
Y_TYPE* y = (Y_TYPE *) malloc(sizeof(Y_TYPE) * y_ne);
float* d = (float *) malloc(sizeof(float) * d_ne);
for (size_t i = 0; i < x_ne; i++) {
if (std::is_same<float, X_TYPE>()) {
x[i] = (rand() / (float)RAND_MAX) * 2.0f - 1.0f;
} else if (std::is_same<ggml_fp16_t, X_TYPE>()) {
x[i] = ggml_fp32_to_fp16((rand() / (float)RAND_MAX) * 2.0f - 1.0f);
} else {
GGML_ASSERT(false);
}
}
for (size_t i = 0; i < y_ne; i++) {
if (std::is_same<float, Y_TYPE>()) {
// y[i] = (rand() / (float)RAND_MAX) * 2.0f - 1.0f;
y[i] = (i % k == i / k) ? 1.0f : 0.0f;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
} else if (std::is_same<ggml_fp16_t, Y_TYPE>()) {
// y[i] = ggml_fp32_to_fp16((rand() / (float)RAND_MAX) * 2.0f - 1.0f);
y[i] = ggml_fp32_to_fp16((i % k == i / k) ? 1.0f : 0.0f);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
} else {
GGML_ASSERT(false);
}
}
ggml_vk_buffer_write(ctx, d_X, 0, x, sizeof(X_TYPE) * k * m * batch);
ggml_vk_buffer_write(ctx, d_Y, 0, y, sizeof(Y_TYPE) * k * n * batch);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
vk_context * subctx = ggml_vk_create_context(ctx, ctx->device->compute_queue);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
for (size_t i = 0; i < num_it; i++) {
ggml_vk_ctx_begin(ctx, subctx);
ggml_vk_matmul(
ctx, subctx, p, ggml_vk_subbuffer(d_X), ggml_vk_subbuffer(d_Y), ggml_vk_subbuffer(d_D), ggml_vk_subbuffer(ctx->prealloc_split_k),
m, n, k,
k, k, m, k*m, k*n, m*n,
split_k, batch, batch, batch, 1, 1
);
ggml_vk_ctx_end(subctx);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
auto begin = std::chrono::high_resolution_clock::now();
ggml_vk_submit(subctx, ctx->fence);
VK_CHECK(ctx->device->device.waitForFences({ ctx->fence }, true, UINT64_MAX), "ggml_vk_test_matmul waitForFences");
ctx->device->device.resetFences({ ctx->fence });
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
auto end = std::chrono::high_resolution_clock::now();
double time = std::chrono::duration_cast<std::chrono::microseconds>(end-begin).count() / 1000.0;
// copy dst to host
ggml_vk_buffer_read(ctx, d_D, 0, d, sizeof(float) * d_ne);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
float * d_chk = (float *) malloc(sizeof(float) * d_ne);
ggml_init_params iparams = {
/*.mem_size =*/ 1024*1024*1024,
/*.mem_buffer =*/ NULL,
/*.no_alloc =*/ true,
};
ggml_context * ggml_ctx = ggml_init(iparams);
ggml_type src0_type;
ggml_type src1_type;
if (std::is_same<float, X_TYPE>()) {
src0_type = GGML_TYPE_F32;
} else if (std::is_same<ggml_fp16_t, X_TYPE>()) {
src0_type = GGML_TYPE_F16;
} else {
GGML_ASSERT(false);
}
if (std::is_same<float, Y_TYPE>()) {
src1_type = GGML_TYPE_F32;
} else if (std::is_same<ggml_fp16_t, Y_TYPE>()) {
src1_type = GGML_TYPE_F16;
} else {
GGML_ASSERT(false);
}
ggml_tensor * src0_ggml = ggml_new_tensor_3d(ggml_ctx, src0_type, k, m, batch);
ggml_tensor * src1_ggml = ggml_new_tensor_3d(ggml_ctx, src1_type, k, n, batch);
ggml_tensor * tensor_ggml = ggml_mul_mat(ggml_ctx, src0_ggml, src1_ggml);
src0_ggml->data = x;
src1_ggml->data = y;
tensor_ggml->data = d_chk;
ggml_cgraph * cgraph = ggml_new_graph(ggml_ctx);
ggml_build_forward_expand(cgraph, tensor_ggml);
ggml_graph_compute_with_ctx(ggml_ctx, cgraph, 1);
ggml_free(ggml_ctx);
double avg_err = 0.0;
int first_err_n = -1;
int first_err_m = -1;
int first_err_b = -1;
for (size_t i = 0; i < m*n*batch; i++) {
double err = std::fabs(d[i] - d_chk[i]);
avg_err += err;
if (err > 0.05f && first_err_n == -1) {
first_err_b = i / (m * n);
first_err_n = (i % (m * n)) / m;
first_err_m = (i % (m * n)) % m;
}
}
avg_err /= m * n;
std::cerr << "TEST " << shname << " m=" << m << " n=" << n << " k=" << k << " batch=" << batch << " split_k=" << split_k << " matmul " << time / num_it << "ms avg_err=" << avg_err << std::endl;
if (avg_err > 0.1) {
std::cerr << "m = " << first_err_m << " n = " << first_err_n << " b = " << first_err_b << std::endl;
std::cerr << "Actual result: " << std::endl << std::endl;
ggml_vk_print_matrix_area(d, GGML_TYPE_F32, m, n, first_err_m, first_err_n, first_err_b);
std::cerr << std::endl;
ggml_vk_print_matrix_area(d, GGML_TYPE_F32, m, n, first_err_m, first_err_n + 15, first_err_b);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
std::cerr << "Expected result: " << std::endl << std::endl;
ggml_vk_print_matrix_area(d_chk, GGML_TYPE_F32, m, n, first_err_m, first_err_n, first_err_b);
if (split_k > 1) {
float * split_k_buf = (float *) malloc(sizeof(float) * d_ne * split_k);
ggml_vk_buffer_read(ctx, ctx->prealloc_split_k, 0, split_k_buf, sizeof(float) * d_ne * split_k);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
std::cerr << "d_buf0: " << std::endl << std::endl;
ggml_vk_print_matrix_area(split_k_buf, GGML_TYPE_F32, m, n, first_err_m, first_err_n, first_err_b);
std::cerr << "d_buf1: " << std::endl << std::endl;
ggml_vk_print_matrix_area(split_k_buf + d_ne, GGML_TYPE_F32, m, n, first_err_m, first_err_n, first_err_b);
std::cerr << "d_buf2: " << std::endl << std::endl;
ggml_vk_print_matrix_area(split_k_buf + 2 * d_ne, GGML_TYPE_F32, m, n, first_err_m, first_err_n, first_err_b);
std::cerr << "d_buf3: " << std::endl << std::endl;
ggml_vk_print_matrix_area(split_k_buf + 3 * d_ne, GGML_TYPE_F32, m, n, first_err_m, first_err_n, first_err_b);
free(split_k_buf);
}
}
free(d_chk);
ggml_vk_queue_cleanup(ctx, ctx->device->transfer_queue);
ggml_vk_queue_cleanup(ctx, ctx->device->compute_queue);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
ggml_vk_destroy_buffer(d_X);
ggml_vk_destroy_buffer(d_Y);
ggml_vk_destroy_buffer(d_D);
ggml_pipeline_cleanup(p);
ggml_pipeline_cleanup(ctx->device->pipeline_matmul_split_k_reduce);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
free(x);
free(y);
free(d);
}
static void ggml_vk_print_tensor_area(const ggml_tensor * tensor, int i0, int i1, int i2, int i3) {
if (tensor->type != GGML_TYPE_F32 && tensor->type != GGML_TYPE_F16) {
return;
}
i0 = std::max(i0, 5);
i1 = std::max(i1, 5);
i2 = std::max(i2, 0);
i3 = std::max(i3, 0);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
fprintf(stderr, " ");
for (int idx1 = i1 - 5; idx1 < i1 + 5; idx1++) {
fprintf(stderr, "%7d ", idx1);
}
fprintf(stderr, "\n");
for (int idx0 = i0 - 5; idx0 < i0 + 5; idx0++) {
fprintf(stderr, "%7d: ", idx0);
for (int idx1 = i1 - 5; idx1 < i1 + 5; idx1++) {
if (idx0 >= 0 && idx0 < tensor->ne[0] && idx1 >= 0 && idx1 < tensor->ne[1] && i2 >= 0 && i2 < tensor->ne[2] && i3 >= 0 && i3 < tensor->ne[3]) {
float val;
if (tensor->type == GGML_TYPE_F32) {
val = *(float *) ((char *) tensor->data + i3*tensor->nb[3] + i2*tensor->nb[2] + idx1*tensor->nb[1] + idx0*tensor->nb[0]);
} else if (tensor->type == GGML_TYPE_F16) {
val = ggml_fp16_to_fp32(*(ggml_fp16_t *) ((char *) tensor->data + i3*tensor->nb[3] + i2*tensor->nb[2] + idx1*tensor->nb[1] + idx0*tensor->nb[0]));
} else {
GGML_ASSERT(false);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
fprintf(stderr, "% 7.2f ", val);
} else {
fprintf(stderr, " ");
}
}
fprintf(stderr, "\n");
}
}
static void ggml_vk_test_transfer(ggml_backend_vk_context * ctx, size_t ne, bool pinned) {
VK_LOG_DEBUG("ggml_vk_test_transfer(" << ne << ")");
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
// Check transfers are correct
vk_buffer buffer = ggml_vk_create_buffer_check(ctx, sizeof(float) * ne, vk::MemoryPropertyFlagBits::eDeviceLocal);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
float * x;
float * y;
if (pinned) {
x = (float *) ggml_vk_host_malloc(ctx, sizeof(float) * ne);
y = (float *) ggml_vk_host_malloc(ctx, sizeof(float) * ne);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
} else {
x = (float *) malloc(sizeof(float) * ne);
y = (float *) malloc(sizeof(float) * ne);
}
for (size_t i = 0; i < ne; i++) {
x[i] = rand() / (float)RAND_MAX;
}
vk_context * subctx = ggml_vk_create_context(ctx, ctx->device->compute_queue);
ggml_vk_ctx_begin(ctx, subctx);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
auto begin = std::chrono::high_resolution_clock::now();
ggml_vk_buffer_write_async(ctx, subctx, buffer, 0, x, sizeof(float) * ne);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
for (auto& cpy : subctx->in_memcpys) {
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
memcpy(cpy.dst, cpy.src, cpy.n);
}
subctx->in_memcpys.clear();
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
ggml_vk_ctx_end(subctx);
ggml_vk_submit(subctx, ctx->fence);
VK_CHECK(ctx->device->device.waitForFences({ ctx->fence }, true, UINT64_MAX), "ggml_vk_test_transfer waitForFences");
ctx->device->device.resetFences({ ctx->fence });
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
auto end = std::chrono::high_resolution_clock::now();
double ms_to_gpu = std::chrono::duration_cast<std::chrono::microseconds>(end-begin).count() / 1000.0;
ggml_vk_ctx_begin(ctx, subctx);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
begin = std::chrono::high_resolution_clock::now();
ggml_vk_buffer_read_async(ctx, subctx, buffer, 0, y, sizeof(float) * ne);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
ggml_vk_ctx_end(subctx);
ggml_vk_submit(subctx, ctx->fence);
VK_CHECK(ctx->device->device.waitForFences({ ctx->fence }, true, UINT64_MAX), "ggml_vk_test_transfer waitForFences");
ctx->device->device.resetFences({ ctx->fence });
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
for (auto& cpy : subctx->out_memcpys) {
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
memcpy(cpy.dst, cpy.src, cpy.n);
}
subctx->out_memcpys.clear();
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
end = std::chrono::high_resolution_clock::now();
double ms_from_gpu = std::chrono::duration_cast<std::chrono::microseconds>(end-begin).count() / 1000.0;
double avg_err = 0.0;
for (size_t i = 0; i < ne; i++) {
avg_err += std::fabs(x[i] - y[i]);
}
double kb = ne * sizeof(float) / 1024.0;
std::cerr << "TEST TRANSFER " << kb << " KB to_gpu " << ms_to_gpu << "ms (" << kb / ms_to_gpu * 1000.0 / 1024.0 << " MB/s) from_gpu " << ms_from_gpu << "ms (" << kb / ms_from_gpu * 1000.0 / 1024.0 << " MB/s) avg_err=" << avg_err / ne << std::endl;
ggml_vk_destroy_buffer(buffer);
if (pinned) {
ggml_vk_host_free(ctx, x);
ggml_vk_host_free(ctx, y);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
} else {
free(x);
free(y);
}
}
static void ggml_vk_quantize_data(const float * from, void * to, size_t ne, ggml_type quant) {
ggml_quantize_chunk(quant, from, to, 0, 1, ne, nullptr);
}
static void ggml_vk_test_dequant(ggml_backend_vk_context * ctx, size_t ne, ggml_type quant) {
VK_LOG_DEBUG("ggml_vk_test_dequant(" << ne << ")");
const size_t x_sz = sizeof(float) * ne;
const size_t x_sz_f16 = sizeof(ggml_fp16_t) * ne;
const size_t qx_sz = ne * ggml_type_size(quant)/ggml_blck_size(quant);
float * x = (float *) malloc(x_sz);
void * qx = malloc(qx_sz);
vk_buffer qx_buf = ggml_vk_create_buffer_check(ctx, qx_sz, vk::MemoryPropertyFlagBits::eDeviceLocal);
vk_buffer x_buf = ggml_vk_create_buffer_check(ctx, x_sz_f16, vk::MemoryPropertyFlagBits::eDeviceLocal);
ggml_fp16_t * x_chk = (ggml_fp16_t *) malloc(x_sz_f16);
for (size_t i = 0; i < ne; i++) {
x[i] = rand() / (float)RAND_MAX;
}
vk_pipeline p = ctx->device->pipeline_dequant[quant];
ggml_vk_quantize_data(x, qx, ne, quant);
ggml_pipeline_allocate_descriptor_sets(ctx, p, 1);
ggml_vk_buffer_write(ctx, qx_buf, 0, qx, qx_sz);
vk_context * subctx = ggml_vk_create_context(ctx, ctx->device->compute_queue);
ggml_vk_ctx_begin(ctx, subctx);
const std::vector<uint32_t> pc = { 1, (uint32_t)ne, (uint32_t)ne, (uint32_t)ne, (uint32_t)ne };
ggml_vk_dispatch_pipeline(ctx, subctx, p, { { qx_buf, 0, qx_sz }, { x_buf, 0, x_sz_f16 } }, pc.size() * sizeof(int), pc.data(), { (uint32_t)ne, 1, 1});
ggml_vk_ctx_end(subctx);
auto begin = std::chrono::high_resolution_clock::now();
ggml_vk_submit(subctx, ctx->fence);
VK_CHECK(ctx->device->device.waitForFences({ ctx->fence }, true, UINT64_MAX), "ggml_vk_test_dequant waitForFences");
ctx->device->device.resetFences({ ctx->fence });
auto end = std::chrono::high_resolution_clock::now();
double ms_dequant = std::chrono::duration_cast<std::chrono::microseconds>(end-begin).count() / 1000.0;
ggml_vk_buffer_read(ctx, x_buf, 0, x_chk, x_sz_f16);
int first_err = -1;
double avg_err = 0.0;
for (size_t i = 0; i < ne; i++) {
double error = std::fabs(x[i] - ggml_fp16_to_fp32(x_chk[i]));
avg_err += error;
if (first_err < 0 && error > 0.05) {
first_err = i;
}
}
avg_err /= ne;
std::cerr << "TEST DEQUANT " << ggml_type_name(quant) << " time=" << ms_dequant << "ms avg_err=" << avg_err << std::endl;
if (avg_err > 0.1) {
std::cerr << "first_error = " << first_err << std::endl;
std::cerr << "Actual result: " << std::endl << std::endl;
for (int i = std::max(0, first_err - 5); i < std::min((int)ne, first_err + 5); i++) {
std::cerr << ggml_fp16_to_fp32(x_chk[i]) << ", ";
}
std::cerr << std::endl << "Expected result: " << std::endl << std::endl;
for (int i = std::max(0, first_err - 5); i < std::min((int)ne, first_err + 5); i++) {
std::cerr << x[i] << ", ";
}
std::cerr << std::endl;
}
ggml_vk_destroy_buffer(x_buf);
ggml_vk_destroy_buffer(qx_buf);
free(x);
free(qx);
free(x_chk);
}
static void ggml_vk_test_dequant_matmul(ggml_backend_vk_context * ctx, size_t m, size_t n, size_t k, size_t batch, size_t num_it, size_t split_k, size_t shader_size, ggml_type quant) {
VK_LOG_DEBUG("ggml_vk_test_dequant_matmul(" << m << ", " << n << ", " << k << ", " << batch << ", " << num_it << ", " << split_k << ", " << ggml_type_name(quant) << ")");
const size_t x_ne = m * k * batch;
const size_t y_ne = k * n * batch;
const size_t d_ne = m * n * batch;
vk_pipeline p;
std::string shname;
if (shader_size == 0) {
p = ctx->device->pipeline_dequant_mul_mat_mat[quant]->a_s;
shname = std::string(ggml_type_name(quant)) + "_ALIGNED_S";
} else if (shader_size == 1) {
p = ctx->device->pipeline_dequant_mul_mat_mat[quant]->a_m;
shname = std::string(ggml_type_name(quant)) + "_ALIGNED_M";
} else if (shader_size == 2) {
p = ctx->device->pipeline_dequant_mul_mat_mat[quant]->a_l;
shname = std::string(ggml_type_name(quant)) + "_ALIGNED_L";
} else {
GGML_ASSERT(0);
}
const size_t kpad = ggml_vk_align_size(k, p->align);
if (k != kpad) {
if (shader_size == 0) {
p = ctx->device->pipeline_dequant_mul_mat_mat[quant]->s;
shname = std::string(ggml_type_name(quant)) + "_S";
} else if (shader_size == 1) {
p = ctx->device->pipeline_dequant_mul_mat_mat[quant]->m;
shname = std::string(ggml_type_name(quant)) + "_M";
} else if (shader_size == 2) {
p = ctx->device->pipeline_dequant_mul_mat_mat[quant]->l;
shname = std::string(ggml_type_name(quant)) + "_L";
} else {
GGML_ASSERT(0);
}
}
const size_t x_sz = sizeof(float) * x_ne;
const size_t y_sz = sizeof(float) * y_ne;
const size_t qx_sz = x_ne * ggml_type_size(quant)/ggml_blck_size(quant);
const size_t d_sz = sizeof(float) * d_ne;
float * x = (float *) malloc(x_sz);
float * y = (float *) malloc(y_sz);
void * qx = malloc(qx_sz);
vk_buffer qx_buf = ggml_vk_create_buffer_check(ctx, qx_sz, vk::MemoryPropertyFlagBits::eDeviceLocal);
vk_buffer y_buf = ggml_vk_create_buffer_check(ctx, y_sz, vk::MemoryPropertyFlagBits::eDeviceLocal);
vk_buffer d_buf = ggml_vk_create_buffer_check(ctx, d_sz, vk::MemoryPropertyFlagBits::eDeviceLocal);
float * d = (float *) malloc(d_sz);
float * d_chk = (float *) malloc(d_sz);
for (size_t i = 0; i < x_ne; i++) {
x[i] = (rand() / (float)RAND_MAX) * 2.0f - 1.0f;
}
ggml_vk_quantize_data(x, qx, x_ne, quant);
for (size_t i = 0; i < y_ne; i++) {
// y[i] = rand() / (float)RAND_MAX;
y[i] = (i % k == i / k) ? 1.0f : 0.0f;
}
ggml_pipeline_allocate_descriptor_sets(ctx, p, num_it);
if (split_k > 1) {
ggml_pipeline_allocate_descriptor_sets(ctx, ctx->device->pipeline_matmul_split_k_reduce, num_it);
if (ctx->prealloc_split_k == nullptr || ctx->prealloc_split_k->size < sizeof(float) * d_ne * split_k) {
// Resize buffer
if (ctx->prealloc_split_k != nullptr) {
ggml_vk_destroy_buffer(ctx->prealloc_split_k);
}
ctx->prealloc_split_k = ggml_vk_create_buffer_check(ctx, sizeof(float) * d_ne * split_k, vk::MemoryPropertyFlagBits::eDeviceLocal);
}
}
ggml_vk_buffer_write(ctx, qx_buf, 0, qx, qx_sz);
ggml_vk_buffer_write(ctx, y_buf, 0, y, y_sz);
vk_context * subctx = ggml_vk_create_context(ctx, ctx->device->compute_queue);
for (size_t i = 0; i < num_it; i++) {
ggml_vk_ctx_begin(ctx, subctx);
ggml_vk_matmul(
ctx, subctx, p, ggml_vk_subbuffer(qx_buf), ggml_vk_subbuffer(y_buf), ggml_vk_subbuffer(d_buf), ggml_vk_subbuffer(ctx->prealloc_split_k),
m, n, k,
k, k, m, k*m, k*n, m*n,
split_k, batch, batch, batch, 1, 1
);
ggml_vk_ctx_end(subctx);
}
auto begin = std::chrono::high_resolution_clock::now();
ggml_vk_submit(subctx, ctx->fence);
VK_CHECK(ctx->device->device.waitForFences({ ctx->fence }, true, UINT64_MAX), "ggml_vk_test_dequant waitForFences");
ctx->device->device.resetFences({ ctx->fence });
auto end = std::chrono::high_resolution_clock::now();
double time_ms = std::chrono::duration_cast<std::chrono::microseconds>(end-begin).count() / 1000.0;
ggml_vk_buffer_read(ctx, d_buf, 0, d, d_sz);
ggml_init_params iparams = {
/*.mem_size =*/ 1024*1024*1024,
/*.mem_buffer =*/ NULL,
/*.no_alloc =*/ true,
};
ggml_context * ggml_ctx = ggml_init(iparams);
ggml_tensor * src0_ggml = ggml_new_tensor_3d(ggml_ctx, quant, k, m, batch);
ggml_tensor * src1_ggml = ggml_new_tensor_3d(ggml_ctx, GGML_TYPE_F32, k, n, batch);
ggml_tensor * tensor_ggml = ggml_mul_mat(ggml_ctx, src0_ggml, src1_ggml);
src0_ggml->data = qx;
src1_ggml->data = y;
tensor_ggml->data = d_chk;
ggml_cgraph * cgraph = ggml_new_graph(ggml_ctx);
ggml_build_forward_expand(cgraph, tensor_ggml);
ggml_graph_compute_with_ctx(ggml_ctx, cgraph, 1);
ggml_free(ggml_ctx);
double avg_err = 0.0;
int first_err_n = -1;
int first_err_m = -1;
int first_err_b = -1;
for (size_t i = 0; i < m*n*batch; i++) {
double err = std::fabs(d[i] - d_chk[i]);
avg_err += err;
if ((err > 0.05f || std::isnan(err)) && first_err_n == -1) {
first_err_b = i / (m * n);
first_err_n = (i % (m * n)) / m;
first_err_m = (i % (m * n)) % m;
}
}
avg_err /= m * n;
std::cerr << "TEST MMQ " << shname << " m=" << m << " n=" << n << " k=" << k << " batch=" << batch << " split_k=" << split_k << " matmul " << time_ms / num_it << "ms avg_err=" << avg_err << std::endl;
if (avg_err > 0.01 || std::isnan(avg_err)) {
std::cerr << "m = " << first_err_m << " n = " << first_err_n << " b = " << first_err_b << std::endl;
std::cerr << "Actual result: " << std::endl << std::endl;
ggml_vk_print_matrix_area(d, GGML_TYPE_F32, m, n, first_err_m, first_err_n, first_err_b);
std::cerr << std::endl;
std::cerr << "Expected result: " << std::endl << std::endl;
ggml_vk_print_matrix_area(d_chk, GGML_TYPE_F32, m, n, first_err_m, first_err_n, first_err_b);
if (split_k > 1) {
float * split_k_buf = (float *) malloc(sizeof(float) * d_ne * split_k);
ggml_vk_buffer_read(ctx, ctx->prealloc_split_k, 0, split_k_buf, sizeof(float) * d_ne * split_k);
std::cerr << "d_buf0: " << std::endl << std::endl;
ggml_vk_print_matrix_area(split_k_buf, GGML_TYPE_F32, m, n, first_err_m, first_err_n, first_err_b);
std::cerr << "d_buf1: " << std::endl << std::endl;
ggml_vk_print_matrix_area(split_k_buf + d_ne, GGML_TYPE_F32, m, n, first_err_m, first_err_n, first_err_b);
std::cerr << "d_buf2: " << std::endl << std::endl;
ggml_vk_print_matrix_area(split_k_buf + 2 * d_ne, GGML_TYPE_F32, m, n, first_err_m, first_err_n, first_err_b);
std::cerr << "d_buf3: " << std::endl << std::endl;
ggml_vk_print_matrix_area(split_k_buf + 3 * d_ne, GGML_TYPE_F32, m, n, first_err_m, first_err_n, first_err_b);
free(split_k_buf);
}
}
ggml_vk_destroy_buffer(qx_buf);
ggml_vk_destroy_buffer(y_buf);
ggml_vk_destroy_buffer(d_buf);
free(x);
free(qx);
free(y);
free(d);
free(d_chk);
}
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
#endif
static ggml_tensor_extra_gpu * ggml_vk_tensor_create_extra(ggml_tensor * tensor) {
VK_LOG_DEBUG("ggml_vk_create_extra(" << tensor << " (" << tensor->name << ", " << ggml_op_name(tensor->op) << "))");
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
ggml_tensor_extra_gpu * extra = new ggml_tensor_extra_gpu;
extra->reset();
tensor->extra = extra;
return extra;
}
static void ggml_vk_preallocate_buffers_graph(ggml_backend_vk_context * ctx, ggml_tensor * node){
VK_LOG_DEBUG("ggml_vk_preallocate_buffers_graph(" << node << ")");
ggml_tensor_extra_gpu * extra = (ggml_tensor_extra_gpu *) node->extra;
if (extra == nullptr) {
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
return;
}
ggml_tensor * src0 = node->src[0];
ggml_tensor * src1 = node->src[1];
const bool use_src0 = src0 != nullptr;
const int64_t ne00 = use_src0 ? src0->ne[0] : 0;
const int64_t ne01 = use_src0 ? src0->ne[1] : 0;
const int64_t ne02 = use_src0 ? src0->ne[2] : 0;
const int64_t ne03 = use_src0 ? src0->ne[3] : 0;
const bool use_src1 = src1 != nullptr && node->op != GGML_OP_CPY && node->op != GGML_OP_CONT && node->op != GGML_OP_DUP;
const int64_t ne10 = use_src1 ? src1->ne[0] : 0;
const int64_t ne11 = use_src1 ? src1->ne[1] : 0;
const int64_t ne12 = use_src1 ? src1->ne[2] : 0;
const int64_t ne13 = use_src1 ? src1->ne[3] : 0;
const int64_t ne20 = node->ne[0];
const int64_t ne21 = node->ne[1];
const int64_t ne22 = node->ne[2];
const int64_t ne23 = node->ne[3];
const ggml_type src0_type = (use_src0 && src0->type == GGML_TYPE_F32) ? src0->type : GGML_TYPE_F16;
const ggml_type src1_type = (use_src1 && src1->type == GGML_TYPE_F32) ? src1->type : GGML_TYPE_F16;
const bool x_non_contig = use_src0 && !ggml_vk_dim01_contiguous(src0);
const bool y_non_contig = use_src1 && !ggml_vk_dim01_contiguous(src1);
const bool y_f32_kernel = use_src1 && src1->type == GGML_TYPE_F32 && !y_non_contig;
bool mmp = (use_src0 && use_src1 && src1_type == GGML_TYPE_F32) ? ggml_vk_get_mul_mat_mat_pipeline(ctx, src0_type, y_non_contig ? GGML_TYPE_F16 : src1->type) != nullptr : false;
const bool qx_needs_dequant = use_src0 && (!mmp || x_non_contig);
const bool qy_needs_dequant = use_src1 && ((src1->type != GGML_TYPE_F16 && !y_f32_kernel) || y_non_contig);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
int split_k;
if (node->op == GGML_OP_MUL_MAT || node->op == GGML_OP_MUL_MAT_ID) {
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
split_k = ggml_vk_guess_split_k(ne01, ne11, ne10);
} else {
split_k = 1;
}
const uint32_t x_ne = ne00 * ne01;
const uint32_t y_ne = ne10 * ne11;
const uint32_t d_ne = ne20 * ne21;
const uint64_t x_sz = (use_src0 && qx_needs_dequant) ? ggml_vk_align_size(sizeof(src0_type) * x_ne, ctx->device->properties.limits.minStorageBufferOffsetAlignment) * ne02 * ne03 : 0;
const uint64_t y_sz = (use_src1 && qy_needs_dequant) ? ggml_vk_align_size(sizeof(src1_type) * y_ne, ctx->device->properties.limits.minStorageBufferOffsetAlignment) * ne12 * ne13 : 0;
uint64_t d_sz = ggml_vk_align_size(ggml_type_size(node->type) * d_ne, ctx->device->properties.limits.minStorageBufferOffsetAlignment) * ne22 * ne23;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
const uint64_t split_k_size = split_k > 1 ? d_sz * 4 : 0;
if (extra->buffer_gpu.expired()) {
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
// Workaround for CPU backend BLAS matmul calls
extra->buffer_gpu = ggml_vk_create_buffer_temp(ctx, d_sz);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
switch (node->op) {
case GGML_OP_REPEAT:
case GGML_OP_GET_ROWS:
case GGML_OP_RESHAPE:
case GGML_OP_VIEW:
case GGML_OP_PERMUTE:
case GGML_OP_TRANSPOSE:
case GGML_OP_ADD:
case GGML_OP_SCALE:
case GGML_OP_SQR:
case GGML_OP_CLAMP:
case GGML_OP_CPY:
case GGML_OP_CONT:
case GGML_OP_DUP:
case GGML_OP_MUL:
case GGML_OP_DIV:
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
case GGML_OP_NORM:
case GGML_OP_RMS_NORM:
case GGML_OP_DIAG_MASK_INF:
case GGML_OP_SOFT_MAX:
case GGML_OP_ROPE:
case GGML_OP_ARGSORT:
case GGML_OP_SUM_ROWS:
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
break;
case GGML_OP_UNARY:
switch (ggml_get_unary_op(node)) {
case GGML_UNARY_OP_SILU:
case GGML_UNARY_OP_GELU:
case GGML_UNARY_OP_RELU:
break;
default:
return;
}
break;
case GGML_OP_MUL_MAT:
case GGML_OP_MUL_MAT_ID:
if (ctx->prealloc_size_x < x_sz) {
ctx->prealloc_size_x = x_sz;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
if (ctx->prealloc_size_y < y_sz) {
ctx->prealloc_size_y = y_sz;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
if (ctx->prealloc_size_split_k < split_k_size) {
ctx->prealloc_size_split_k = split_k_size;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
if (ctx->staging_size < x_sz + y_sz) {
ctx->staging_size = x_sz + y_sz;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
break;
default:
return;
}
}
static void ggml_vk_preallocate_buffers(ggml_backend_vk_context * ctx) {
#if defined(GGML_VULKAN_RUN_TESTS)
ctx->staging = ggml_vk_create_buffer_check(ctx, 100ul * 1024ul * 1024ul,
vk::MemoryPropertyFlagBits::eHostVisible | vk::MemoryPropertyFlagBits::eHostCoherent | vk::MemoryPropertyFlagBits::eHostCached,
vk::MemoryPropertyFlagBits::eHostVisible | vk::MemoryPropertyFlagBits::eHostCoherent);
ggml_vk_test_transfer(ctx, 8192 * 1000, false);
ggml_vk_test_transfer(ctx, 8192 * 1000, true);
ggml_vk_test_dequant(ctx, 7680, GGML_TYPE_F32);
ggml_vk_test_dequant(ctx, 7680, GGML_TYPE_Q4_0);
ggml_vk_test_dequant(ctx, 7680, GGML_TYPE_Q4_1);
ggml_vk_test_dequant(ctx, 7680, GGML_TYPE_Q5_0);
ggml_vk_test_dequant(ctx, 7680, GGML_TYPE_Q5_1);
ggml_vk_test_dequant(ctx, 7680, GGML_TYPE_Q8_0);
ggml_vk_test_dequant(ctx, 7680, GGML_TYPE_Q2_K);
ggml_vk_test_dequant(ctx, 7680, GGML_TYPE_Q3_K);
ggml_vk_test_dequant(ctx, 7680, GGML_TYPE_Q4_K);
ggml_vk_test_dequant(ctx, 7680, GGML_TYPE_Q5_K);
ggml_vk_test_dequant(ctx, 7680, GGML_TYPE_Q6_K);
ggml_vk_test_matmul<ggml_fp16_t, ggml_fp16_t>(ctx, 512, 512, 100, 32, 100, 1, 2);
ggml_vk_test_matmul<float, float>(ctx, 128, 512, 512, 2, 100, 1, 0);
ggml_vk_test_matmul<float, float>(ctx, 128, 512, 512, 2, 100, 1, 1);
ggml_vk_test_matmul<float, float>(ctx, 128, 512, 512, 2, 100, 1, 2);
ggml_vk_test_matmul<float, float>(ctx, 128, 512, 512, 2, 100, 4, 0);
ggml_vk_test_matmul<float, float>(ctx, 128, 512, 512, 2, 100, 4, 1);
ggml_vk_test_matmul<float, float>(ctx, 128, 512, 512, 2, 100, 4, 2);
ggml_vk_test_dequant_matmul(ctx, 128, 512, 512, 2, 100, 1, 0, GGML_TYPE_Q4_0);
ggml_vk_test_dequant_matmul(ctx, 128, 512, 512, 2, 100, 1, 1, GGML_TYPE_Q4_0);
ggml_vk_test_dequant_matmul(ctx, 128, 512, 512, 2, 100, 1, 2, GGML_TYPE_Q4_0);
ggml_vk_test_dequant_matmul(ctx, 128, 512, 512, 2, 100, 4, 0, GGML_TYPE_Q4_0);
ggml_vk_test_dequant_matmul(ctx, 128, 512, 512, 2, 100, 4, 1, GGML_TYPE_Q4_0);
ggml_vk_test_dequant_matmul(ctx, 128, 512, 512, 2, 100, 4, 2, GGML_TYPE_Q4_0);
ggml_vk_test_dequant_matmul(ctx, 128, 512, 512, 2, 100, 1, 0, GGML_TYPE_Q4_1);
ggml_vk_test_dequant_matmul(ctx, 128, 512, 512, 2, 100, 1, 1, GGML_TYPE_Q4_1);
ggml_vk_test_dequant_matmul(ctx, 128, 512, 512, 2, 100, 1, 2, GGML_TYPE_Q4_1);
ggml_vk_test_dequant_matmul(ctx, 128, 512, 512, 2, 100, 4, 0, GGML_TYPE_Q4_1);
ggml_vk_test_dequant_matmul(ctx, 128, 512, 512, 2, 100, 4, 1, GGML_TYPE_Q4_1);
ggml_vk_test_dequant_matmul(ctx, 128, 512, 512, 2, 100, 4, 2, GGML_TYPE_Q4_1);
ggml_vk_test_dequant_matmul(ctx, 128, 512, 512, 2, 100, 1, 0, GGML_TYPE_Q5_0);
ggml_vk_test_dequant_matmul(ctx, 128, 512, 512, 2, 100, 1, 1, GGML_TYPE_Q5_0);
ggml_vk_test_dequant_matmul(ctx, 128, 512, 512, 2, 100, 1, 2, GGML_TYPE_Q5_0);
ggml_vk_test_dequant_matmul(ctx, 128, 512, 512, 2, 100, 4, 0, GGML_TYPE_Q5_0);
ggml_vk_test_dequant_matmul(ctx, 128, 512, 512, 2, 100, 4, 1, GGML_TYPE_Q5_0);
ggml_vk_test_dequant_matmul(ctx, 128, 512, 512, 2, 100, 4, 2, GGML_TYPE_Q5_0);
ggml_vk_test_dequant_matmul(ctx, 128, 512, 512, 2, 100, 1, 0, GGML_TYPE_Q5_1);
ggml_vk_test_dequant_matmul(ctx, 128, 512, 512, 2, 100, 1, 1, GGML_TYPE_Q5_1);
ggml_vk_test_dequant_matmul(ctx, 128, 512, 512, 2, 100, 1, 2, GGML_TYPE_Q5_1);
ggml_vk_test_dequant_matmul(ctx, 128, 512, 512, 2, 100, 4, 0, GGML_TYPE_Q5_1);
ggml_vk_test_dequant_matmul(ctx, 128, 512, 512, 2, 100, 4, 1, GGML_TYPE_Q5_1);
ggml_vk_test_dequant_matmul(ctx, 128, 512, 512, 2, 100, 4, 2, GGML_TYPE_Q5_1);
ggml_vk_test_dequant_matmul(ctx, 128, 512, 512, 2, 100, 1, 0, GGML_TYPE_Q8_0);
ggml_vk_test_dequant_matmul(ctx, 128, 512, 512, 2, 100, 1, 1, GGML_TYPE_Q8_0);
ggml_vk_test_dequant_matmul(ctx, 128, 512, 512, 2, 100, 1, 2, GGML_TYPE_Q8_0);
ggml_vk_test_dequant_matmul(ctx, 128, 512, 512, 2, 100, 4, 0, GGML_TYPE_Q8_0);
ggml_vk_test_dequant_matmul(ctx, 128, 512, 512, 2, 100, 4, 1, GGML_TYPE_Q8_0);
ggml_vk_test_dequant_matmul(ctx, 128, 512, 512, 2, 100, 4, 2, GGML_TYPE_Q8_0);
ggml_vk_test_dequant_matmul(ctx, 128, 512, 512, 2, 100, 1, 0, GGML_TYPE_Q2_K);
ggml_vk_test_dequant_matmul(ctx, 128, 512, 512, 2, 100, 1, 1, GGML_TYPE_Q2_K);
ggml_vk_test_dequant_matmul(ctx, 128, 512, 512, 2, 100, 1, 2, GGML_TYPE_Q2_K);
ggml_vk_test_dequant_matmul(ctx, 128, 512, 512, 2, 100, 4, 0, GGML_TYPE_Q2_K);
ggml_vk_test_dequant_matmul(ctx, 128, 512, 512, 2, 100, 4, 1, GGML_TYPE_Q2_K);
ggml_vk_test_dequant_matmul(ctx, 128, 512, 512, 2, 100, 4, 2, GGML_TYPE_Q2_K);
ggml_vk_test_dequant_matmul(ctx, 128, 512, 512, 2, 100, 1, 0, GGML_TYPE_Q3_K);
ggml_vk_test_dequant_matmul(ctx, 128, 512, 512, 2, 100, 1, 1, GGML_TYPE_Q3_K);
ggml_vk_test_dequant_matmul(ctx, 128, 512, 512, 2, 100, 1, 2, GGML_TYPE_Q3_K);
ggml_vk_test_dequant_matmul(ctx, 128, 512, 512, 2, 100, 4, 0, GGML_TYPE_Q3_K);
ggml_vk_test_dequant_matmul(ctx, 128, 512, 512, 2, 100, 4, 1, GGML_TYPE_Q3_K);
ggml_vk_test_dequant_matmul(ctx, 128, 512, 512, 2, 100, 4, 2, GGML_TYPE_Q3_K);
ggml_vk_test_dequant_matmul(ctx, 128, 512, 512, 2, 100, 1, 0, GGML_TYPE_Q4_K);
ggml_vk_test_dequant_matmul(ctx, 128, 512, 512, 2, 100, 1, 1, GGML_TYPE_Q4_K);
ggml_vk_test_dequant_matmul(ctx, 128, 512, 512, 2, 100, 1, 2, GGML_TYPE_Q4_K);
ggml_vk_test_dequant_matmul(ctx, 128, 512, 512, 2, 100, 4, 0, GGML_TYPE_Q4_K);
ggml_vk_test_dequant_matmul(ctx, 128, 512, 512, 2, 100, 4, 1, GGML_TYPE_Q4_K);
ggml_vk_test_dequant_matmul(ctx, 128, 512, 512, 2, 100, 4, 2, GGML_TYPE_Q4_K);
ggml_vk_test_dequant_matmul(ctx, 128, 512, 512, 2, 100, 1, 0, GGML_TYPE_Q5_K);
ggml_vk_test_dequant_matmul(ctx, 128, 512, 512, 2, 100, 1, 1, GGML_TYPE_Q5_K);
ggml_vk_test_dequant_matmul(ctx, 128, 512, 512, 2, 100, 1, 2, GGML_TYPE_Q5_K);
ggml_vk_test_dequant_matmul(ctx, 128, 512, 512, 2, 100, 4, 0, GGML_TYPE_Q5_K);
ggml_vk_test_dequant_matmul(ctx, 128, 512, 512, 2, 100, 4, 1, GGML_TYPE_Q5_K);
ggml_vk_test_dequant_matmul(ctx, 128, 512, 512, 2, 100, 4, 2, GGML_TYPE_Q5_K);
ggml_vk_test_dequant_matmul(ctx, 128, 512, 512, 2, 100, 1, 0, GGML_TYPE_Q6_K);
ggml_vk_test_dequant_matmul(ctx, 128, 512, 512, 2, 100, 1, 1, GGML_TYPE_Q6_K);
ggml_vk_test_dequant_matmul(ctx, 128, 512, 512, 2, 100, 1, 2, GGML_TYPE_Q6_K);
ggml_vk_test_dequant_matmul(ctx, 128, 512, 512, 2, 100, 4, 0, GGML_TYPE_Q6_K);
ggml_vk_test_dequant_matmul(ctx, 128, 512, 512, 2, 100, 4, 1, GGML_TYPE_Q6_K);
ggml_vk_test_dequant_matmul(ctx, 128, 512, 512, 2, 100, 4, 2, GGML_TYPE_Q6_K);
std::cerr << std::endl;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
const std::vector<size_t> vals {
8, 8, 8,
100, 46, 576,
623, 111, 128,
100, 46, 558,
512, 1, 256,
128, 110, 622,
511, 511, 127,
511, 511, 7,
511, 511, 17,
49, 49, 128,
128, 49, 49,
4096, 49, 4096,
11008, 49, 4096,
4096, 49, 11008,
32000, 49, 4096,
512, 512, 128,
128, 512, 512,
4096, 512, 4096,
11008, 512, 4096,
4096, 512, 11008,
32000, 512, 4096,
};
const size_t num_it = 1;
for (size_t i = 0; i < vals.size(); i += 3) {
ggml_vk_test_matmul<ggml_fp16_t, float>(ctx, vals[i], vals[i + 1], vals[i + 2], 2, num_it, 1, 0);
ggml_vk_test_matmul<ggml_fp16_t, float>(ctx, vals[i], vals[i + 1], vals[i + 2], 2, num_it, 1, 1);
ggml_vk_test_matmul<ggml_fp16_t, float>(ctx, vals[i], vals[i + 1], vals[i + 2], 2, num_it, 1, 2);
ggml_vk_test_matmul<ggml_fp16_t, float>(ctx, vals[i], vals[i + 1], vals[i + 2], 2, num_it, 4, 0);
ggml_vk_test_matmul<ggml_fp16_t, float>(ctx, vals[i], vals[i + 1], vals[i + 2], 2, num_it, 4, 1);
ggml_vk_test_matmul<ggml_fp16_t, float>(ctx, vals[i], vals[i + 1], vals[i + 2], 2, num_it, 4, 2);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
std::cerr << std::endl;
}
GGML_ASSERT(false);
#endif
if (ctx->prealloc_x == nullptr || (ctx->prealloc_size_x > 0 && ctx->prealloc_x->size < ctx->prealloc_size_x)) {
VK_LOG_MEMORY("ggml_vk_preallocate_buffers(x_size: " << ctx->prealloc_size_x << ")");
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
// Resize buffer
if (ctx->prealloc_x != nullptr) {
ggml_vk_destroy_buffer(ctx->prealloc_x);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
ctx->prealloc_x = ggml_vk_create_buffer_device(ctx->device, ctx->prealloc_size_x);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
if (ctx->prealloc_y == nullptr || (ctx->prealloc_size_y > 0 && ctx->prealloc_y->size < ctx->prealloc_size_y)) {
VK_LOG_MEMORY("ggml_vk_preallocate_buffers(y_size: " << ctx->prealloc_size_y << ")");
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
// Resize buffer
if (ctx->prealloc_y != nullptr) {
ggml_vk_destroy_buffer(ctx->prealloc_y);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
ctx->prealloc_y = ggml_vk_create_buffer_device(ctx->device, ctx->prealloc_size_y);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
if (ctx->prealloc_split_k == nullptr || (ctx->prealloc_size_split_k > 0 && ctx->prealloc_split_k->size < ctx->prealloc_size_split_k)) {
VK_LOG_MEMORY("ggml_vk_preallocate_buffers(split_k_size: " << ctx->prealloc_size_split_k << ")");
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
// Resize buffer
if (ctx->prealloc_split_k != nullptr) {
ggml_vk_destroy_buffer(ctx->prealloc_split_k);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
ctx->prealloc_split_k = ggml_vk_create_buffer_device(ctx->device, ctx->prealloc_size_split_k);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
if (ctx->staging == nullptr || (ctx->staging_size > 0 && ctx->staging->size < ctx->staging_size)) {
VK_LOG_MEMORY("ggml_vk_preallocate_buffers(staging_size: " << ctx->staging_size << ")");
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
// Resize buffer
if (ctx->staging != nullptr) {
ggml_vk_destroy_buffer(ctx->staging);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
ctx->staging = ggml_vk_create_buffer_check(ctx->device, ctx->staging_size,
vk::MemoryPropertyFlagBits::eHostVisible | vk::MemoryPropertyFlagBits::eHostCoherent | vk::MemoryPropertyFlagBits::eHostCached,
vk::MemoryPropertyFlagBits::eHostVisible | vk::MemoryPropertyFlagBits::eHostCoherent);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
}
static void ggml_vk_build_graph(ggml_backend_vk_context * ctx, ggml_tensor * node, bool last_node){
ggml_tensor_extra_gpu * extra = (ggml_tensor_extra_gpu *) node->extra;
if (ggml_is_empty(node) || extra == nullptr) {
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
return;
}
VK_LOG_DEBUG("ggml_vk_build_graph(" << node << ", " << ggml_op_name(node->op) << ")");
ctx->semaphore_idx = 0;
ctx->staging_offset = 0;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
const ggml_tensor * src0 = node->src[0];
const ggml_tensor * src1 = node->src[1];
const ggml_tensor * src2 = node->src[2];
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
switch (node->op) {
// Return on empty ops to avoid generating a compute_ctx and setting exit_tensor
case GGML_OP_RESHAPE:
case GGML_OP_VIEW:
case GGML_OP_PERMUTE:
case GGML_OP_TRANSPOSE:
case GGML_OP_NONE:
return;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
case GGML_OP_UNARY:
switch (ggml_get_unary_op(node)) {
case GGML_UNARY_OP_SILU:
case GGML_UNARY_OP_GELU:
case GGML_UNARY_OP_RELU:
break;
default:
return;
}
break;
case GGML_OP_REPEAT:
case GGML_OP_GET_ROWS:
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
case GGML_OP_ADD:
case GGML_OP_MUL:
case GGML_OP_DIV:
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
case GGML_OP_SCALE:
case GGML_OP_SQR:
case GGML_OP_CLAMP:
case GGML_OP_CPY:
case GGML_OP_CONT:
case GGML_OP_DUP:
case GGML_OP_NORM:
case GGML_OP_RMS_NORM:
case GGML_OP_DIAG_MASK_INF:
case GGML_OP_SOFT_MAX:
case GGML_OP_ROPE:
case GGML_OP_MUL_MAT:
case GGML_OP_MUL_MAT_ID:
case GGML_OP_ARGSORT:
case GGML_OP_SUM_ROWS:
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
break;
default:
std::cerr << "ggml_vulkan: Error: Missing op: " << ggml_op_name(node->op) << std::endl;
GGML_ASSERT(false);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
return;
}
if (ctx->compute_ctx == nullptr) {
ctx->compute_ctx = ggml_vk_create_context(ctx, ctx->device->compute_queue);
ggml_vk_ctx_begin(ctx->device, ctx->compute_ctx);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
switch (node->op) {
case GGML_OP_REPEAT:
ggml_vk_repeat(ctx, ctx->compute_ctx, src0, src1, node);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
break;
case GGML_OP_GET_ROWS:
ggml_vk_get_rows(ctx, ctx->compute_ctx, src0, src1, node);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
break;
case GGML_OP_ADD:
ggml_vk_add(ctx, ctx->compute_ctx, src0, src1, node);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
break;
case GGML_OP_MUL:
ggml_vk_mul(ctx, ctx->compute_ctx, src0, src1, node);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
break;
case GGML_OP_DIV:
ggml_vk_div(ctx, ctx->compute_ctx, src0, src1, node);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
break;
case GGML_OP_SCALE:
ggml_vk_scale(ctx, ctx->compute_ctx, src0, node);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
break;
case GGML_OP_SQR:
ggml_vk_sqr(ctx, ctx->compute_ctx, src0, node);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
break;
case GGML_OP_CLAMP:
ggml_vk_clamp(ctx, ctx->compute_ctx, src0, node);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
break;
case GGML_OP_CPY:
case GGML_OP_CONT:
case GGML_OP_DUP:
ggml_vk_cpy(ctx, ctx->compute_ctx, src0, node);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
break;
case GGML_OP_NORM:
ggml_vk_norm(ctx, ctx->compute_ctx, src0, node);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
break;
case GGML_OP_RMS_NORM:
ggml_vk_rms_norm(ctx, ctx->compute_ctx, src0, node);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
break;
case GGML_OP_UNARY:
switch (ggml_get_unary_op(node)) {
case GGML_UNARY_OP_SILU:
case GGML_UNARY_OP_GELU:
case GGML_UNARY_OP_RELU:
ggml_vk_unary(ctx, ctx->compute_ctx, src0, node);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
break;
default:
return;
}
break;
case GGML_OP_DIAG_MASK_INF:
ggml_vk_diag_mask_inf(ctx, ctx->compute_ctx, src0, node);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
break;
case GGML_OP_SOFT_MAX:
ggml_vk_soft_max(ctx, ctx->compute_ctx, src0, src1, node);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
break;
case GGML_OP_ROPE:
ggml_vk_rope(ctx, ctx->compute_ctx, src0, src1, src2, node);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
break;
case GGML_OP_ARGSORT:
ggml_vk_argsort(ctx, ctx->compute_ctx, src0, node);
break;
case GGML_OP_SUM_ROWS:
ggml_vk_sum_rows(ctx, ctx->compute_ctx, src0, node);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
break;
case GGML_OP_MUL_MAT:
ggml_vk_mul_mat(ctx, ctx->compute_ctx, src0, src1, node);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
break;
case GGML_OP_MUL_MAT_ID:
ggml_vk_mul_mat_id(ctx, ctx->compute_ctx, src0, src1, src2, node);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
break;
default:
return;
}
extra->ctx_idx = ctx->compute_ctx->idx;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
#ifdef GGML_VULKAN_CHECK_RESULTS
// Force context reset on each node so that each tensor ends up in its own context
// and can be run and compared to its CPU equivalent separately
last_node = true;
#endif
if (last_node) {
ggml_vk_ctx_end(ctx->compute_ctx);
ctx->compute_ctx->exit_tensor = node;
ctx->compute_ctx = nullptr;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
}
static bool ggml_vk_compute_forward(ggml_backend_vk_context * ctx, ggml_tensor * tensor){
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
ggml_tensor_extra_gpu * extra = nullptr;
switch (tensor->op) {
case GGML_OP_ADD:
case GGML_OP_GET_ROWS:
case GGML_OP_MUL:
case GGML_OP_DIV:
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
case GGML_OP_SCALE:
case GGML_OP_SQR:
case GGML_OP_CLAMP:
case GGML_OP_CPY:
case GGML_OP_CONT:
case GGML_OP_DUP:
case GGML_OP_NORM:
case GGML_OP_RMS_NORM:
case GGML_OP_DIAG_MASK_INF:
case GGML_OP_SOFT_MAX:
case GGML_OP_ROPE:
case GGML_OP_RESHAPE:
case GGML_OP_VIEW:
case GGML_OP_PERMUTE:
case GGML_OP_TRANSPOSE:
case GGML_OP_NONE:
case GGML_OP_ARGSORT:
case GGML_OP_SUM_ROWS:
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
extra = (ggml_tensor_extra_gpu *) tensor->extra;
break;
case GGML_OP_UNARY:
switch (ggml_get_unary_op(tensor)) {
case GGML_UNARY_OP_SILU:
case GGML_UNARY_OP_GELU:
case GGML_UNARY_OP_RELU:
extra = (ggml_tensor_extra_gpu *) tensor->extra;
break;
default:
return false;
}
break;
case GGML_OP_MUL_MAT:
case GGML_OP_MUL_MAT_ID:
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
extra = (ggml_tensor_extra_gpu *) tensor->extra;
break;
default:
return false;
}
if (extra == nullptr) {
return false;
}
VK_LOG_DEBUG("ggml_vk_compute_forward(" << tensor << ", name=" << tensor->name << ", op=" << ggml_op_name(tensor->op) << ", type=" << tensor->type << ", ne0=" << tensor->ne[0] << ", ne1=" << tensor->ne[1] << ", ne2=" << tensor->ne[2] << ", ne3=" << tensor->ne[3] << ", nb0=" << tensor->nb[0] << ", nb1=" << tensor->nb[1] << ", nb2=" << tensor->nb[2] << ", nb3=" << tensor->nb[3] << ", view_src=" << tensor->view_src << ", view_offs=" << tensor->view_offs << ")");
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
#ifdef GGML_VULKAN_CHECK_RESULTS
ggml_vk_check_results_0(ctx, tensor);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
#endif
vk_context& subctx = ctx->gc.contexts[extra->ctx_idx];
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
// Only run if ctx hasn't been submitted yet
if (!subctx.seqs.empty()) {
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
// Do staging buffer copies
for (auto& cpy : subctx.in_memcpys) {
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
memcpy(cpy.dst, cpy.src, cpy.n);
}
ggml_vk_submit(&subctx, ctx->fence);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
if (tensor == subctx.exit_tensor) {
VK_CHECK(ctx->device->device.waitForFences({ ctx->fence }, true, UINT64_MAX), "ggml_vk_compute_forward waitForFences");
ctx->device->device.resetFences({ ctx->fence });
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
// Do staging buffer copies
for (auto& cpy : subctx.out_memcpys) {
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
memcpy(cpy.dst, cpy.src, cpy.n);
}
subctx.in_memcpys.clear();
subctx.out_memcpys.clear();
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
return true;
}
// Clean up after graph processing is done
static void ggml_vk_graph_cleanup(ggml_backend_vk_context * ctx) {
VK_LOG_DEBUG("ggml_vk_graph_cleanup()");
for (auto& buffer : ctx->gc.temp_buffers) {
ggml_vk_pool_free(ctx, buffer);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
ctx->gc.temp_buffers.clear();
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
for (auto& pipeline : ctx->device->pipelines) {
if (pipeline.expired()) {
continue;
}
vk_pipeline pl = pipeline.lock();
ggml_pipeline_cleanup(pl);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
ggml_vk_queue_cleanup(ctx->device, ctx->device->compute_queue);
ggml_vk_queue_cleanup(ctx->device, ctx->device->transfer_queue);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
for (size_t i = 0; i < ctx->gc.semaphores.size(); i++) {
ctx->device->device.destroySemaphore({ ctx->gc.semaphores[i].s });
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
ctx->gc.semaphores.clear();
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
for (size_t i = 0; i < ctx->gc.tl_semaphores.size(); i++) {
ctx->device->device.destroySemaphore({ ctx->gc.tl_semaphores[i].s });
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
ctx->gc.tl_semaphores.clear();
ctx->semaphore_idx = 0;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
ctx->event_idx = 0;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
for (auto& event : ctx->gc.events) {
ctx->device->device.resetEvent(event);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
ctx->staging_offset = 0;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
ctx->compute_ctx = nullptr;
ctx->transfer_ctx = nullptr;
ctx->gc.contexts.clear();
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
// Clean up on backend free
static void ggml_vk_cleanup(ggml_backend_vk_context * ctx) {
VK_LOG_DEBUG("ggml_vk_cleanup(" << ctx->name << ")");
ggml_vk_graph_cleanup(ctx);
ggml_vk_destroy_buffer(ctx->prealloc_x);
ggml_vk_destroy_buffer(ctx->prealloc_y);
ggml_vk_destroy_buffer(ctx->prealloc_split_k);
ggml_vk_destroy_buffer(ctx->staging);
for (auto& buffer : ctx->buffer_pool) {
ggml_vk_destroy_buffer(buffer);
}
ctx->prealloc_size_x = 0;
ctx->prealloc_size_y = 0;
ctx->prealloc_size_split_k = 0;
ctx->staging_size = 0;
for (auto& event : ctx->gc.events) {
ctx->device->device.destroyEvent(event);
}
ctx->gc.events.clear();
ctx->device->device.destroyFence(ctx->fence);
}
GGML_CALL static int ggml_vk_get_device_count() {
ggml_vk_instance_init();
return vk_instance.device_indices.size();
}
GGML_CALL static void ggml_vk_get_device_description(int device, char * description, size_t description_size) {
ggml_vk_instance_init();
std::vector<vk::PhysicalDevice> devices = vk_instance.instance.enumeratePhysicalDevices();
vk::PhysicalDeviceProperties props;
devices[device].getProperties(&props);
snprintf(description, description_size, "%s", props.deviceName.data());
}
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
// backend interface
#define UNUSED GGML_UNUSED
// device backend
static void * const vk_ptr_base = (void *)(uintptr_t) 0x1000; // NOLINT
struct ggml_backend_vk_buffer_context {
vk_device_ref device;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
vk_buffer dev_buffer;
ggml_tensor_extra_gpu * temp_tensor_extras = nullptr;
size_t temp_tensor_extra_index = 0;
std::string name;
ggml_backend_vk_buffer_context(vk_device_ref device, vk_buffer&& dev_buffer, std::string& name) :
device(device),
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
dev_buffer(dev_buffer),
name(name) {
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
~ggml_backend_vk_buffer_context() {
ggml_vk_destroy_buffer(dev_buffer);
if (temp_tensor_extras != nullptr) {
delete[] temp_tensor_extras;
}
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
ggml_tensor_extra_gpu * ggml_vk_alloc_temp_tensor_extra() {
if (temp_tensor_extras == nullptr) {
temp_tensor_extras = new ggml_tensor_extra_gpu[GGML_VK_MAX_NODES];
}
size_t alloc_index = temp_tensor_extra_index;
temp_tensor_extra_index = (temp_tensor_extra_index + 1) % GGML_VK_MAX_NODES;
ggml_tensor_extra_gpu * extra = &temp_tensor_extras[alloc_index];
extra->reset();
return extra;
}
};
GGML_CALL static const char * ggml_backend_vk_buffer_get_name(ggml_backend_buffer_t buffer) {
ggml_backend_vk_buffer_context * ctx = (ggml_backend_vk_buffer_context *)buffer->context;
return ctx->name.c_str();
}
GGML_CALL static bool ggml_backend_buffer_is_vk(ggml_backend_buffer_t buffer) {
return buffer->iface.get_name == ggml_backend_vk_buffer_get_name;
}
GGML_CALL static void ggml_backend_vk_buffer_free_buffer(ggml_backend_buffer_t buffer) {
VK_LOG_MEMORY("ggml_backend_vk_buffer_free_buffer()");
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
ggml_backend_vk_buffer_context * ctx = (ggml_backend_vk_buffer_context *)buffer->context;
ggml_vk_destroy_buffer(ctx->dev_buffer);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
delete ctx;
}
GGML_CALL static void * ggml_backend_vk_buffer_get_base(ggml_backend_buffer_t buffer) {
return vk_ptr_base;
UNUSED(buffer);
}
GGML_CALL static void ggml_backend_vk_buffer_init_tensor(ggml_backend_buffer_t buffer, ggml_tensor * tensor) {
VK_LOG_DEBUG("ggml_backend_vk_buffer_init_tensor(" << buffer << " (" << buffer->context << "), " << tensor << ")");
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
ggml_backend_vk_buffer_context * ctx = (ggml_backend_vk_buffer_context *)buffer->context;
if (tensor->view_src != nullptr) {
GGML_ASSERT(tensor->view_src->buffer->buft == buffer->buft);
GGML_ASSERT(tensor->view_src->extra != nullptr);
tensor->extra = tensor->view_src->extra;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
} else {
ggml_tensor_extra_gpu * extra = ctx->ggml_vk_alloc_temp_tensor_extra();
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
extra->buffer_gpu = ctx->dev_buffer;
extra->offset = (uint8_t *) tensor->data - (uint8_t *) vk_ptr_base;
tensor->extra = extra;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
}
GGML_CALL static void ggml_backend_vk_buffer_set_tensor(ggml_backend_buffer_t buffer, ggml_tensor * tensor, const void * data, size_t offset, size_t size) {
VK_LOG_DEBUG("ggml_backend_vk_buffer_set_tensor(" << buffer << ", " << tensor << ", " << data << ", " << offset << ", " << size << ")");
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
ggml_tensor_extra_gpu * extra = (ggml_tensor_extra_gpu *) tensor->extra;
vk_buffer buf = extra->buffer_gpu.lock();
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
ggml_vk_buffer_write(buf, extra->offset + tensor->view_offs + offset, data, size);
GGML_UNUSED(buffer);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
GGML_CALL static void ggml_backend_vk_buffer_get_tensor(ggml_backend_buffer_t buffer, const ggml_tensor * tensor, void * data, size_t offset, size_t size) {
VK_LOG_DEBUG("ggml_backend_vk_buffer_get_tensor(" << buffer << ", " << tensor << ", " << data << ", " << offset << ", " << size << ")");
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
ggml_tensor_extra_gpu * extra = (ggml_tensor_extra_gpu *) tensor->extra;
vk_buffer buf = extra->buffer_gpu.lock();
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
ggml_vk_buffer_read(buf, extra->offset + tensor->view_offs + offset, data, size);
GGML_UNUSED(buffer);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
GGML_CALL static bool ggml_backend_vk_buffer_cpy_tensor(ggml_backend_buffer_t buffer, const ggml_tensor * src, ggml_tensor * dst) {
if (ggml_backend_buffer_is_vk(src->buffer)) {
ggml_tensor_extra_gpu * src_extra = (ggml_tensor_extra_gpu *) src->extra;
ggml_tensor_extra_gpu * dst_extra = (ggml_tensor_extra_gpu *) dst->extra;
vk_buffer src_buf = src_extra->buffer_gpu.lock();
vk_buffer dst_buf = dst_extra->buffer_gpu.lock();
ggml_vk_buffer_copy(dst_buf, dst_extra->offset + dst->view_offs, src_buf, src_extra->offset + src->view_offs, ggml_nbytes(src));
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
return true;
}
return false;
UNUSED(buffer);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
GGML_CALL static void ggml_backend_vk_buffer_clear(ggml_backend_buffer_t buffer, uint8_t value) {
ggml_backend_vk_buffer_context * ctx = (ggml_backend_vk_buffer_context *)buffer->context;
ggml_vk_buffer_memset(ctx->dev_buffer, 0, value, buffer->size);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
static ggml_backend_buffer_i ggml_backend_vk_buffer_interface = {
/* .get_name = */ ggml_backend_vk_buffer_get_name,
/* .free_buffer = */ ggml_backend_vk_buffer_free_buffer,
/* .get_base = */ ggml_backend_vk_buffer_get_base,
/* .init_tensor = */ ggml_backend_vk_buffer_init_tensor,
/* .set_tensor = */ ggml_backend_vk_buffer_set_tensor,
/* .get_tensor = */ ggml_backend_vk_buffer_get_tensor,
/* .cpy_tensor = */ ggml_backend_vk_buffer_cpy_tensor,
/* .clear = */ ggml_backend_vk_buffer_clear,
/* .reset = */ NULL,
};
// vk buffer type
GGML_CALL static const char * ggml_backend_vk_buffer_type_name(ggml_backend_buffer_type_t buft) {
ggml_backend_vk_buffer_type_context * ctx = (ggml_backend_vk_buffer_type_context *)buft->context;
return ctx->name.c_str();
}
GGML_CALL static ggml_backend_buffer_t ggml_backend_vk_buffer_type_alloc_buffer(ggml_backend_buffer_type_t buft, size_t size) {
VK_LOG_MEMORY("ggml_backend_vk_buffer_type_alloc_buffer(" << size << ")");
ggml_backend_vk_buffer_type_context * ctx = (ggml_backend_vk_buffer_type_context *) buft->context;
vk_buffer dev_buffer = nullptr;
try {
dev_buffer = ggml_vk_create_buffer_device(ctx->device, size);
} catch (const vk::SystemError& e) {
return nullptr;
}
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
ggml_backend_vk_buffer_context * bufctx = new ggml_backend_vk_buffer_context(ctx->device, std::move(dev_buffer), ctx->name);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
return ggml_backend_buffer_init(buft, ggml_backend_vk_buffer_interface, bufctx, size);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
GGML_CALL static size_t ggml_backend_vk_buffer_type_get_alignment(ggml_backend_buffer_type_t buft) {
ggml_backend_vk_buffer_type_context * ctx = (ggml_backend_vk_buffer_type_context *) buft->context;
return ctx->device->properties.limits.minStorageBufferOffsetAlignment;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
GGML_CALL static size_t ggml_backend_vk_buffer_type_get_max_size(ggml_backend_buffer_type_t buft) {
ggml_backend_vk_buffer_type_context * ctx = (ggml_backend_vk_buffer_type_context *) buft->context;
return ctx->device->max_memory_allocation_size;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
GGML_CALL static size_t ggml_backend_vk_buffer_type_get_alloc_size(ggml_backend_buffer_type_t buft, const ggml_tensor * tensor) {
return ggml_nbytes(tensor);
UNUSED(buft);
}
GGML_CALL ggml_backend_buffer_type_t ggml_backend_vk_buffer_type(size_t dev_num) {
ggml_vk_instance_init();
VK_LOG_DEBUG("ggml_backend_vk_buffer_type(" << dev_num << ")");
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
vk_device dev = ggml_vk_get_device(dev_num);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
return &dev->buffer_type;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
// host buffer type
GGML_CALL static const char * ggml_backend_vk_host_buffer_type_name(ggml_backend_buffer_type_t buft) {
return GGML_VK_NAME "_Host";
UNUSED(buft);
}
GGML_CALL static const char * ggml_backend_vk_host_buffer_name(ggml_backend_buffer_t buffer) {
return GGML_VK_NAME "_Host";
UNUSED(buffer);
}
GGML_CALL static void ggml_backend_vk_host_buffer_free_buffer(ggml_backend_buffer_t buffer) {
VK_LOG_MEMORY("ggml_backend_vk_host_buffer_free_buffer()");
ggml_vk_host_free(vk_instance.devices[0], buffer->context);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
GGML_CALL static ggml_backend_buffer_t ggml_backend_vk_host_buffer_type_alloc_buffer(ggml_backend_buffer_type_t buft, size_t size) {
VK_LOG_MEMORY("ggml_backend_vk_host_buffer_type_alloc_buffer(" << size << ")");
size += 32; // Behave like the CPU buffer type
void * ptr = nullptr;
try {
ptr = ggml_vk_host_malloc(vk_instance.devices[0], size);
} catch (vk::SystemError& e) {
std::cerr << "ggml_vulkan: Failed to allocate pinned memory." << std::endl;
std::cerr << "ggml_vulkan: " << e.what() << std::endl;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
// fallback to cpu buffer
return ggml_backend_buft_alloc_buffer(ggml_backend_cpu_buffer_type(), size);
}
ggml_backend_buffer_t buffer = ggml_backend_cpu_buffer_from_ptr(ptr, size);
buffer->buft = buft;
buffer->iface.get_name = ggml_backend_vk_host_buffer_name;
buffer->iface.free_buffer = ggml_backend_vk_host_buffer_free_buffer;
return buffer;
UNUSED(buft);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
GGML_CALL static size_t ggml_backend_vk_host_buffer_type_get_alignment(ggml_backend_buffer_type_t buft) {
return vk_instance.devices[0]->properties.limits.minMemoryMapAlignment;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
UNUSED(buft);
}
// Should be changed to return device-specific host buffer type
// but that probably requires changes in llama.cpp
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
GGML_CALL ggml_backend_buffer_type_t ggml_backend_vk_host_buffer_type() {
static struct ggml_backend_buffer_type ggml_backend_vk_buffer_type_host = {
/* .iface = */ {
/* .get_name = */ ggml_backend_vk_host_buffer_type_name,
/* .alloc_buffer = */ ggml_backend_vk_host_buffer_type_alloc_buffer,
/* .get_alignment = */ ggml_backend_vk_host_buffer_type_get_alignment,
/* .get_max_size = */ NULL, // defaults to SIZE_MAX
/* .get_alloc_size = */ ggml_backend_cpu_buffer_type()->iface.get_alloc_size,
/* .is_host = */ ggml_backend_cpu_buffer_type()->iface.is_host,
},
/* .context = */ nullptr,
};
// Make sure device 0 is initialized
ggml_vk_instance_init();
ggml_vk_get_device(0);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
return &ggml_backend_vk_buffer_type_host;
}
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
// backend
GGML_CALL static const char * ggml_backend_vk_name(ggml_backend_t backend) {
ggml_backend_vk_context * ctx = (ggml_backend_vk_context *)backend->context;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
return ctx->name.c_str();
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
GGML_CALL static void ggml_backend_vk_free(ggml_backend_t backend) {
ggml_backend_vk_context * ctx = (ggml_backend_vk_context *)backend->context;
VK_LOG_DEBUG("ggml_backend_vk_free(" << ctx->name << ")");
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
ggml_vk_cleanup(ctx);
delete ctx;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
delete backend;
}
GGML_CALL static ggml_backend_buffer_type_t ggml_backend_vk_get_default_buffer_type(ggml_backend_t backend) {
ggml_backend_vk_context * ctx = (ggml_backend_vk_context *)backend->context;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
return &ctx->device->buffer_type;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
GGML_CALL static void ggml_backend_vk_set_tensor_async(ggml_backend_t backend, ggml_tensor * tensor, const void * data, size_t offset, size_t size) {
VK_LOG_DEBUG("ggml_backend_vk_set_tensor_async(" << size << ")");
ggml_backend_vk_context * ctx = (ggml_backend_vk_context *)backend->context;
GGML_ASSERT((tensor->buffer->buft == ggml_backend_vk_get_default_buffer_type(backend) || tensor->buffer->buft == ggml_backend_vk_host_buffer_type()) && "unsupported buffer type");
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
ggml_tensor_extra_gpu * extra = (ggml_tensor_extra_gpu *) tensor->extra;
if (ctx->transfer_ctx == nullptr) {
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
// Initialize new transfer context
ctx->transfer_ctx = ggml_vk_create_context(ctx, ctx->device->transfer_queue);
ggml_vk_ctx_begin(ctx->device, ctx->transfer_ctx);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
vk_buffer buf = extra->buffer_gpu.lock();
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
ggml_vk_buffer_write_async(ctx->transfer_ctx, buf, extra->offset + tensor->view_offs + offset, data, size, ctx->staging, ctx->staging_offset);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
GGML_CALL static void ggml_backend_vk_get_tensor_async(ggml_backend_t backend, const ggml_tensor * tensor, void * data, size_t offset, size_t size) {
VK_LOG_DEBUG("ggml_backend_vk_get_tensor_async(" << size << ")");
ggml_backend_vk_context * ctx = (ggml_backend_vk_context *)backend->context;
GGML_ASSERT((tensor->buffer->buft == ggml_backend_vk_get_default_buffer_type(backend) || tensor->buffer->buft == ggml_backend_vk_host_buffer_type()) && "unsupported buffer type");
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
ggml_tensor_extra_gpu * extra = (ggml_tensor_extra_gpu *) tensor->extra;
if (ctx->transfer_ctx == nullptr) {
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
// Initialize new transfer context
ctx->transfer_ctx = ggml_vk_create_context(ctx, ctx->device->transfer_queue);
ggml_vk_ctx_begin(ctx->device, ctx->transfer_ctx);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
vk_buffer buf = extra->buffer_gpu.lock();
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
ggml_vk_buffer_read_async(ctx->transfer_ctx, buf, extra->offset + tensor->view_offs + offset, data, size, ctx->staging, ctx->staging_offset);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
GGML_CALL static bool ggml_backend_vk_cpy_tensor_async(ggml_backend_t backend, const ggml_tensor * src, ggml_tensor * dst) {
VK_LOG_DEBUG("ggml_backend_vk_cpy_tensor_async()");
ggml_backend_vk_context * ctx = (ggml_backend_vk_context *)backend->context;
if ((dst->buffer->buft == ggml_backend_vk_get_default_buffer_type(backend) || dst->buffer->buft == ggml_backend_vk_host_buffer_type()) && ggml_backend_buffer_is_vk(src->buffer)) {
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
ggml_tensor_extra_gpu * src_extra = (ggml_tensor_extra_gpu *) src->extra;
ggml_tensor_extra_gpu * dst_extra = (ggml_tensor_extra_gpu *) dst->extra;
if (ctx->transfer_ctx == nullptr) {
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
// Initialize new transfer context
ctx->transfer_ctx = ggml_vk_create_context(ctx, ctx->device->transfer_queue);
ggml_vk_ctx_begin(ctx->device, ctx->transfer_ctx);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
vk_buffer src_buf = src_extra->buffer_gpu.lock();
vk_buffer dst_buf = dst_extra->buffer_gpu.lock();
ggml_vk_buffer_copy_async(ctx->transfer_ctx, dst_buf, dst_extra->offset + dst->view_offs, src_buf, src_extra->offset + src->view_offs, ggml_nbytes(src));
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
return true;
}
return false;
}
GGML_CALL static void ggml_backend_vk_synchronize(ggml_backend_t backend) {
VK_LOG_DEBUG("ggml_backend_vk_synchronize()");
ggml_backend_vk_context * ctx = (ggml_backend_vk_context *)backend->context;
if(ctx->transfer_ctx == nullptr) {
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
return;
}
ggml_vk_ctx_end(ctx->transfer_ctx);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
for (auto& cpy : ctx->transfer_ctx->in_memcpys) {
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
memcpy(cpy.dst, cpy.src, cpy.n);
}
ggml_vk_submit(ctx->transfer_ctx, ctx->fence);
VK_CHECK(ctx->device->device.waitForFences({ ctx->fence }, true, UINT64_MAX), "ggml_backend_vk_synchronize waitForFences");
ctx->device->device.resetFences({ ctx->fence });
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
for (auto& cpy : ctx->transfer_ctx->out_memcpys) {
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
memcpy(cpy.dst, cpy.src, cpy.n);
}
ctx->transfer_ctx = nullptr;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
static bool ggml_vk_is_empty(ggml_tensor * node) {
return ggml_is_empty(node) || node->op == GGML_OP_NONE || node->op == GGML_OP_RESHAPE || node->op == GGML_OP_TRANSPOSE || node->op == GGML_OP_VIEW || node->op == GGML_OP_PERMUTE;
}
GGML_CALL static ggml_status ggml_backend_vk_graph_compute(ggml_backend_t backend, ggml_cgraph * cgraph) {
VK_LOG_DEBUG("ggml_backend_vk_graph_compute(" << cgraph->n_nodes << " nodes)");
ggml_backend_vk_context * ctx = (ggml_backend_vk_context *)backend->context;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
for (int i = 0; i < cgraph->n_nodes; i++) {
ggml_vk_preallocate_buffers_graph(ctx, cgraph->nodes[i]);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
ggml_vk_preallocate_buffers(ctx);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
int last_node = cgraph->n_nodes - 1;
// If the last op in the cgraph isn't backend GPU, the command buffer doesn't get closed properly
while (last_node > 0 && ggml_vk_is_empty(cgraph->nodes[last_node])) {
last_node -= 1;
}
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
for (int i = 0; i < cgraph->n_nodes; i++) {
ggml_vk_build_graph(ctx,cgraph->nodes[i], i == last_node);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
for (int i = 0; i < cgraph->n_nodes; i++) {
ggml_tensor * node = cgraph->nodes[i];
if (ggml_vk_is_empty(node)) {
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
continue;
}
bool ok = ggml_vk_compute_forward(ctx, node);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
if (!ok) {
fprintf(stderr, "%s: error: op not supported %s (%s)\n", __func__, node->name, ggml_op_name(node->op));
}
#ifdef GGML_VULKAN_CHECK_RESULTS
else {
ggml_vk_check_results_1(ctx, node);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
#endif
GGML_ASSERT(ok);
}
ggml_vk_graph_cleanup(ctx);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
return GGML_STATUS_SUCCESS;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
UNUSED(backend);
}
GGML_CALL static bool ggml_backend_vk_supports_op(ggml_backend_t backend, const ggml_tensor * op) {
// ggml_backend_vk_context * ctx = (ggml_backend_vk_context *) backend->context;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
switch (op->op) {
case GGML_OP_UNARY:
switch (ggml_get_unary_op(op)) {
case GGML_UNARY_OP_GELU:
case GGML_UNARY_OP_SILU:
case GGML_UNARY_OP_RELU:
return ggml_is_contiguous(op->src[0]);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
default:
return false;
}
break;
case GGML_OP_MUL_MAT:
case GGML_OP_MUL_MAT_ID:
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
{
switch (op->src[0]->type) {
case GGML_TYPE_F32:
case GGML_TYPE_F16:
case GGML_TYPE_Q4_0:
case GGML_TYPE_Q4_1:
case GGML_TYPE_Q5_0:
case GGML_TYPE_Q5_1:
case GGML_TYPE_Q8_0:
case GGML_TYPE_Q2_K:
case GGML_TYPE_Q3_K:
case GGML_TYPE_Q4_K:
case GGML_TYPE_Q5_K:
case GGML_TYPE_Q6_K:
break;
default:
return false;
}
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
struct ggml_tensor * a;
struct ggml_tensor * b;
if (op->op == GGML_OP_MUL_MAT) {
a = op->src[0];
b = op->src[1];
} else {
a = op->src[2];
b = op->src[1];
}
if (a->ne[3] != b->ne[3]) {
return false;
}
return true;
} break;
case GGML_OP_GET_ROWS:
{
switch (op->src[0]->type) {
case GGML_TYPE_F32:
case GGML_TYPE_F16:
case GGML_TYPE_Q4_0:
case GGML_TYPE_Q4_1:
case GGML_TYPE_Q5_0:
case GGML_TYPE_Q5_1:
case GGML_TYPE_Q8_0:
return true;
default:
return false;
}
} break;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
case GGML_OP_CPY:
case GGML_OP_DUP:
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
{
ggml_type src0_type = op->src[0]->type;
ggml_type src1_type = op->src[1] != nullptr ? op->src[1]->type : src0_type;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
if (src0_type == GGML_TYPE_F32 && src1_type == GGML_TYPE_F32) {
return true;
}
if (src0_type == GGML_TYPE_F32 && src1_type == GGML_TYPE_F16) {
return true;
}
if (src0_type == GGML_TYPE_F16 && src1_type == GGML_TYPE_F16) {
return true;
}
return false;
} break;
// case GGML_OP_REPEAT:
// {
// ggml_type src0_type = op->src[0]->type;
// return src0_type != GGML_TYPE_I32 && src0_type != GGML_TYPE_I16;
// } break;
case GGML_OP_ROPE:
return ggml_is_contiguous(op->src[0]);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
case GGML_OP_NONE:
case GGML_OP_RESHAPE:
case GGML_OP_VIEW:
case GGML_OP_PERMUTE:
case GGML_OP_TRANSPOSE:
case GGML_OP_NORM:
case GGML_OP_ADD:
case GGML_OP_MUL:
case GGML_OP_DIV:
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
case GGML_OP_RMS_NORM:
case GGML_OP_SCALE:
case GGML_OP_SQR:
case GGML_OP_CLAMP:
case GGML_OP_CONT:
case GGML_OP_DIAG_MASK_INF:
case GGML_OP_SOFT_MAX:
case GGML_OP_ARGSORT:
case GGML_OP_SUM_ROWS:
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
return true;
default:
return false;
}
UNUSED(backend);
}
GGML_CALL static bool ggml_backend_vk_offload_op(ggml_backend_t backend, const ggml_tensor * op) {
const int min_batch_size = 32;
return (op->ne[1] >= min_batch_size && op->op != GGML_OP_GET_ROWS) ||
(op->ne[2] >= min_batch_size && op->op == GGML_OP_MUL_MAT_ID);
UNUSED(backend);
}
GGML_CALL static bool ggml_backend_vk_supports_buft(ggml_backend_t backend, ggml_backend_buffer_type_t buft) {
if (buft->iface.get_name != ggml_backend_vk_buffer_type_name) {
return false;
}
ggml_backend_vk_buffer_type_context * buft_ctx = (ggml_backend_vk_buffer_type_context *)buft->context;
ggml_backend_vk_context * ctx = (ggml_backend_vk_context *)backend->context;
return buft_ctx->device == ctx->device;
}
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
// TODO: enable async and synchronize
static ggml_backend_i ggml_backend_vk_interface = {
/* .get_name = */ ggml_backend_vk_name,
/* .free = */ ggml_backend_vk_free,
/* .get_default_buffer_type = */ ggml_backend_vk_get_default_buffer_type,
/* .set_tensor_async = */ NULL, // ggml_backend_vk_set_tensor_async,
/* .get_tensor_async = */ NULL, // ggml_backend_vk_get_tensor_async,
/* .cpy_tensor_async = */ NULL, // ggml_backend_vk_cpy_tensor_async,
/* .synchronize = */ NULL, // ggml_backend_vk_synchronize,
/* .graph_plan_create = */ NULL,
/* .graph_plan_free = */ NULL,
/* .graph_plan_update = */ NULL,
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
/* .graph_plan_compute = */ NULL,
/* .graph_compute = */ ggml_backend_vk_graph_compute,
/* .supports_op = */ ggml_backend_vk_supports_op,
/* .supports_buft = */ ggml_backend_vk_supports_buft,
/* .offload_op = */ ggml_backend_vk_offload_op,
/* .event_new = */ NULL,
/* .event_free = */ NULL,
/* .event_record = */ NULL,
/* .event_wait = */ NULL,
/* .event_synchronize = */ NULL,
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
};
static ggml_guid_t ggml_backend_vk_guid() {
static ggml_guid guid = { 0xb8, 0xf7, 0x4f, 0x86, 0x40, 0x3c, 0xe1, 0x02, 0x91, 0xc8, 0xdd, 0xe9, 0x02, 0x3f, 0xc0, 0x2b };
return &guid;
}
GGML_CALL ggml_backend_t ggml_backend_vk_init(size_t dev_num) {
VK_LOG_DEBUG("ggml_backend_vk_init(" << dev_num << ")");
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
ggml_backend_vk_context * ctx = new ggml_backend_vk_context;
ggml_vk_init(ctx, dev_num);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
ggml_backend_t vk_backend = new ggml_backend {
/* .guid = */ ggml_backend_vk_guid(),
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
/* .interface = */ ggml_backend_vk_interface,
/* .context = */ ctx,
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
};
return vk_backend;
}
GGML_CALL bool ggml_backend_is_vk(ggml_backend_t backend) {
return backend != NULL && ggml_guid_matches(backend->guid, ggml_backend_vk_guid());
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
GGML_CALL int ggml_backend_vk_get_device_count() {
return ggml_vk_get_device_count();
}
GGML_CALL void ggml_backend_vk_get_device_description(int device, char * description, size_t description_size) {
ggml_vk_get_device_description(device, description, description_size);
}
GGML_CALL void ggml_backend_vk_get_device_memory(int device, size_t * free, size_t * total) {
GGML_ASSERT(device < (int) vk_instance.device_indices.size());
vk::PhysicalDevice vkdev = vk_instance.instance.enumeratePhysicalDevices()[vk_instance.device_indices[device]];
vk::PhysicalDeviceMemoryProperties memprops = vkdev.getMemoryProperties();
for (const vk::MemoryHeap& heap : memprops.memoryHeaps) {
if (heap.flags & vk::MemoryHeapFlagBits::eDeviceLocal) {
*total = heap.size;
*free = heap.size;
break;
}
}
}
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
// backend registry
GGML_CALL static ggml_backend_t ggml_backend_reg_vk_init(const char * params, void * user_data) {
ggml_backend_t vk_backend = ggml_backend_vk_init((int) (intptr_t) user_data);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
return vk_backend;
UNUSED(params);
}
extern "C" GGML_CALL int ggml_backend_vk_reg_devices();
GGML_CALL int ggml_backend_vk_reg_devices() {
ggml_vk_instance_init();
for (size_t i = 0; i < vk_instance.device_indices.size(); i++) {
char name[128];
snprintf(name, sizeof(name), "%s%ld", GGML_VK_NAME, i);
ggml_backend_register(name, ggml_backend_reg_vk_init, ggml_backend_vk_buffer_type(i), (void *) (intptr_t) i); // NOLINT
}
return vk_instance.device_indices.size();
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
// Extension availability
static bool ggml_vk_instance_validation_ext_available(const std::vector<vk::ExtensionProperties>& instance_extensions) {
#ifdef GGML_VULKAN_VALIDATE
bool portability_enumeration_ext = false;
// Check for portability enumeration extension for MoltenVK support
for (const auto& properties : instance_extensions) {
if (strcmp("VK_KHR_portability_enumeration", properties.extensionName) == 0) {
return true;
}
}
if (!portability_enumeration_ext) {
std::cerr << "ggml_vulkan: WARNING: Instance extension VK_KHR_portability_enumeration not found." << std::endl;
}
#endif
return false;
UNUSED(instance_extensions);
}
static bool ggml_vk_instance_portability_enumeration_ext_available(const std::vector<vk::ExtensionProperties>& instance_extensions) {
#ifdef __APPLE__
bool portability_enumeration_ext = false;
// Check for portability enumeration extension for MoltenVK support
for (const auto& properties : instance_extensions) {
if (strcmp("VK_KHR_portability_enumeration", properties.extensionName) == 0) {
return true;
}
}
if (!portability_enumeration_ext) {
std::cerr << "ggml_vulkan: WARNING: Instance extension VK_KHR_portability_enumeration not found." << std::endl;
}
#endif
return false;
UNUSED(instance_extensions);
}
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
// checks
#ifdef GGML_VULKAN_CHECK_RESULTS
static void ggml_vk_print_graph_origin(const ggml_tensor * tensor, std::vector<const ggml_tensor *>& done, int level = 0) {
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
if (std::find(done.begin(), done.end(), tensor) != done.end() || level > 10) {
return;
}
for (int j = 0; j < level; j++) {
std::cerr << " ";
}
std::cerr << ggml_op_name(tensor->op) << " gpu=" << (tensor->extra != nullptr) << std::endl;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
done.push_back(tensor);
for (int i = 0; i < GGML_MAX_SRC; i++) {
if (tensor->src[i] != nullptr) {
ggml_vk_print_graph_origin(tensor->src[i], done, level + 1);
}
}
}
static void ggml_vk_print_tensor_area(const ggml_tensor * tensor, const void * data, int i0, int i1, int i2, int i3) {
if (tensor->type != GGML_TYPE_F32 && tensor->type != GGML_TYPE_F16 && tensor->type != GGML_TYPE_I32) {
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
return;
}
i0 = std::max(i0, 5);
i1 = std::max(i1, 5);
i2 = std::max(i2, 0);
i3 = std::max(i3, 0);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
fprintf(stderr, " ");
for (int idx1 = i1 - 5; idx1 < i1 + 5; idx1++) {
fprintf(stderr, "%7d ", idx1);
}
fprintf(stderr, "\n");
for (int idx0 = i0 - 5; idx0 < i0 + 5; idx0++) {
fprintf(stderr, "%7d: ", idx0);
for (int idx1 = i1 - 5; idx1 < i1 + 5; idx1++) {
if (idx0 >= 0 && idx0 < tensor->ne[0] && idx1 >= 0 && idx1 < tensor->ne[1] && i2 >= 0 && i2 < tensor->ne[2] && i3 >= 0 && i3 < tensor->ne[3]) {
float val;
if (tensor->type == GGML_TYPE_F32) {
val = *(const float *) ((const char *) data + i3*tensor->nb[3] + i2*tensor->nb[2] + idx1*tensor->nb[1] + idx0*tensor->nb[0]);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
} else if (tensor->type == GGML_TYPE_F16) {
val = ggml_fp16_to_fp32(*(const ggml_fp16_t *) ((const char *) data + i3*tensor->nb[3] + i2*tensor->nb[2] + idx1*tensor->nb[1] + idx0*tensor->nb[0]));
} else if (tensor->type == GGML_TYPE_I32) {
val = *(const int32_t *) ((const char *) data + i3*tensor->nb[3] + i2*tensor->nb[2] + idx1*tensor->nb[1] + idx0*tensor->nb[0]);
} else {
GGML_ASSERT(false);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
fprintf(stderr, "% 7.2f ", val);
} else {
fprintf(stderr, " ");
}
}
fprintf(stderr, "\n");
}
}
static void ggml_vk_print_tensor(ggml_backend_vk_context * ctx, const ggml_tensor * tensor, const char * name) {
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
void * tensor_data = tensor->data;
if (ggml_backend_buffer_is_vk(tensor->buffer)) {
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
const size_t tensor_size = ggml_nbytes(tensor);
tensor_data = malloc(tensor_size);
ggml_tensor_extra_gpu * extra = (ggml_tensor_extra_gpu *) tensor->extra;
vk_buffer buffer_gpu = extra->buffer_gpu.lock();
ggml_vk_buffer_read(ctx, buffer_gpu, extra->offset + tensor->view_offs, tensor_data, tensor_size);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
std::cerr << "TENSOR CHECK " << name << " (" << tensor->name << "): " << ggml_op_name(tensor->op) << std::endl;
std::cerr << "tensor=" << tensor << " tensor->type: " << ggml_type_name(tensor->type) << " ne0=" << tensor->ne[0] << " nb0=" << tensor->nb[0] << " ne1=" << tensor->ne[1] << " nb1=" << tensor->nb[1] << " ne2=" << tensor->ne[2] << " nb2=" << tensor->nb[2] << " ne3=" << tensor->ne[3] << " nb3=" << tensor->nb[3] << std::endl;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
if (tensor->src[0] != nullptr) {
std::cerr << "tensor->src[0]=" << tensor->src[0] << " name=" << tensor->src[0]->name << " op=" << ggml_op_name(tensor->src[0]->op) << " type=" << ggml_type_name(tensor->src[0]->type) << " ne0=" << tensor->src[0]->ne[0] << " nb0=" << tensor->src[0]->nb[0] << " ne1=" << tensor->src[0]->ne[1] << " nb1=" << tensor->src[0]->nb[1] << " ne2=" << tensor->src[0]->ne[2] << " nb2=" << tensor->src[0]->nb[2] << " ne3=" << tensor->src[0]->ne[3] << " nb3=" << tensor->src[0]->nb[3] << std::endl;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
if (tensor->src[1] != nullptr) {
std::cerr << "tensor->src[1]=" << tensor->src[1] << " name=" << tensor->src[1]->name << " op=" << ggml_op_name(tensor->src[1]->op) << " type=" << ggml_type_name(tensor->src[1]->type) << " ne0=" << tensor->src[1]->ne[0] << " nb0=" << tensor->src[1]->nb[0] << " ne1=" << tensor->src[1]->ne[1] << " nb1=" << tensor->src[1]->nb[1] << " ne2=" << tensor->src[1]->ne[2] << " nb2=" << tensor->src[1]->nb[2] << " ne3=" << tensor->src[1]->ne[3] << " nb3=" << tensor->src[1]->nb[3] << std::endl;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
std::cerr << std::endl << "Result:" << std::endl;
ggml_vk_print_tensor_area(tensor, tensor_data, 5, 5, 0, 0);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
std::cerr << std::endl;
std::cerr << std::endl << "Result:" << std::endl;
ggml_vk_print_tensor_area(tensor, tensor_data, 5, 5, 1, 0);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
std::cerr << std::endl;
std::vector<const ggml_tensor *> done;
ggml_vk_print_graph_origin(tensor, done);
if (ggml_backend_buffer_is_vk(tensor->buffer)) {
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
free(tensor_data);
}
}
void * comp_result;
size_t comp_size;
size_t comp_nb[GGML_MAX_DIMS];
size_t check_counter = 0;
static void ggml_vk_check_results_0(ggml_backend_vk_context * ctx, ggml_tensor * tensor) {
if (tensor->op == GGML_OP_TRANSPOSE) {
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
return;
}
check_counter++;
if (!(vk_output_tensor > 0 && vk_output_tensor == check_counter) && check_counter <= vk_skip_checks) {
return;
}
VK_LOG_DEBUG("ggml_vk_check_results_0(" << tensor->name << ")");
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
ggml_tensor * src0 = tensor->src[0];
ggml_tensor * src1 = tensor->src[1];
ggml_tensor * src2 = tensor->src[2];
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
struct ggml_init_params iparams = {
/*.mem_size =*/ 1024*1024*1024,
/*.mem_buffer =*/ NULL,
/*.no_alloc =*/ false,
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
};
struct ggml_context * ggml_ctx = ggml_init(iparams);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
struct ggml_tensor * src0_clone = nullptr;
struct ggml_tensor * src1_clone = nullptr;
struct ggml_tensor * src2_clone = nullptr;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
struct ggml_tensor * tensor_clone = nullptr;
size_t src0_size;
size_t src1_size;
size_t src2_size;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
void * src0_buffer = nullptr;
void * src1_buffer = nullptr;
void * src2_buffer = nullptr;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
if (src0 != nullptr) {
src0_clone = ggml_dup_tensor(ggml_ctx, src0);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
src0_size = ggml_nbytes(src0);
src0_buffer = malloc(src0_size);
src0_clone->data = src0_buffer;
if (ggml_backend_buffer_is_host(src0->buffer)) {
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
memcpy(src0_clone->data, src0->data, src0_size);
memcpy(src0_clone->nb, src0->nb, sizeof(size_t) * GGML_MAX_DIMS);
} else if (ggml_backend_buffer_is_vk(src0->buffer)) {
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
ggml_tensor_extra_gpu * extra = (ggml_tensor_extra_gpu *) src0->extra;
vk_buffer buffer_gpu = extra->buffer_gpu.lock();
uint64_t offset = extra->offset + src0->view_offs;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
if (!ggml_is_contiguous(src0) && ggml_vk_dim01_contiguous(src0)) {
for (int i3 = 0; i3 < src0->ne[3]; i3++) {
for (int i2 = 0; i2 < src0->ne[2]; i2++) {
const int idx = i3*src0->ne[2] + i2;
ggml_vk_buffer_read(ctx, buffer_gpu, offset + idx * src0->nb[2], ((char *)src0_clone->data + idx * src0_clone->nb[2]), src0->ne[1] * src0->nb[1]);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
}
src0_clone->nb[0] = src0->nb[0];
src0_clone->nb[1] = src0->nb[1];
for (int i = 2; i < GGML_MAX_DIMS; i++) {
src0_clone->nb[i] = src0_clone->nb[i - 1]*src0_clone->ne[i - 1];
}
} else {
if (offset + src0_size >= buffer_gpu->size) {
src0_size = buffer_gpu->size - offset;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
ggml_vk_buffer_read(ctx, buffer_gpu, offset, src0_clone->data, src0_size);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
memcpy(src0_clone->nb, src0->nb, sizeof(size_t) * GGML_MAX_DIMS);
}
} else {
GGML_ASSERT(false);
}
if (vk_output_tensor > 0 && vk_output_tensor == check_counter) {
ggml_vk_print_tensor(ctx, src0, "src0");
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
}
if (src1 != nullptr) {
src1_clone = ggml_dup_tensor(ggml_ctx, src1);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
src1_size = ggml_nbytes(src1);
src1_buffer = malloc(src1_size);
src1_clone->data = src1_buffer;
if (ggml_backend_buffer_is_host(src1->buffer)) {
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
memcpy(src1_clone->data, src1->data, src1_size);
memcpy(src1_clone->nb, src1->nb, sizeof(size_t) * GGML_MAX_DIMS);
} else if (ggml_backend_buffer_is_vk(src1->buffer)) {
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
ggml_tensor_extra_gpu * extra = (ggml_tensor_extra_gpu *) src1->extra;
vk_buffer buffer_gpu = extra->buffer_gpu.lock();
uint64_t offset = extra->offset + src1->view_offs;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
if (!ggml_is_contiguous(src1) && ggml_vk_dim01_contiguous(src1)) {
for (int i3 = 0; i3 < src1->ne[3]; i3++) {
for (int i2 = 0; i2 < src1->ne[2]; i2++) {
const int idx = i3*src1->ne[2] + i2;
ggml_vk_buffer_read(ctx, buffer_gpu, offset + idx * src1->nb[2], ((char *)src1_clone->data + idx * src1_clone->nb[2]), src1->ne[1] * src1->nb[1]);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
}
src1_clone->nb[0] = src1->nb[0];
src1_clone->nb[1] = src1->nb[1];
for (int i = 2; i < GGML_MAX_DIMS; i++) {
src1_clone->nb[i] = src1_clone->nb[i - 1]*src1_clone->ne[i - 1];
}
} else {
if (offset + src1_size >= buffer_gpu->size) {
src1_size = buffer_gpu->size - offset;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
ggml_vk_buffer_read(ctx, buffer_gpu, offset, src1_clone->data, src1_size);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
memcpy(src1_clone->nb, src1->nb, sizeof(size_t) * GGML_MAX_DIMS);
}
} else {
GGML_ASSERT(false);
}
if (vk_output_tensor > 0 && vk_output_tensor == check_counter) {
ggml_vk_print_tensor(ctx, src1, "src1");
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
std::cerr << "TENSOR CHECK: " << ggml_op_name(src1_clone->op) << " (check " << check_counter << ")" << std::endl;
std::cerr << "src1_clone=" << tensor << " src1_clone->type: " << ggml_type_name(src1_clone->type) << " ne0=" << src1_clone->ne[0] << " nb0=" << src1_clone->nb[0] << " ne1=" << src1_clone->ne[1] << " nb1=" << src1_clone->nb[1] << " ne2=" << src1_clone->ne[2] << " nb2=" << src1_clone->nb[2] << " ne3=" << src1_clone->ne[3] << " nb3=" << src1_clone->nb[3] << std::endl;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
if (src1->src[0] != nullptr) {
std::cerr << "src1->src[0]=" << src1->src[0] << " op=" << ggml_op_name(src1->src[0]->op) << " type=" << ggml_type_name(src1->src[0]->type) << " ne0=" << src1->src[0]->ne[0] << " nb0=" << src1->src[0]->nb[0] << " ne1=" << src1->src[0]->ne[1] << " nb1=" << src1->src[0]->nb[1] << " ne2=" << src1->src[0]->ne[2] << " nb2=" << src1->src[0]->nb[2] << " ne3=" << src1->src[0]->ne[3] << " nb3=" << src1->src[0]->nb[3] << std::endl;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
if (src1->src[1] != nullptr) {
std::cerr << "src1->src[1]=" << src1->src[1] << " op=" << ggml_op_name(src1->src[1]->op) << " type=" << ggml_type_name(src1->src[1]->type) << " ne0=" << src1->src[1]->ne[0] << " nb0=" << src1->src[1]->nb[0] << " ne1=" << src1->src[1]->ne[1] << " nb1=" << src1->src[1]->nb[1] << " ne2=" << src1->src[1]->ne[2] << " nb2=" << src1->src[1]->nb[2] << " ne3=" << src1->src[1]->ne[3] << " nb3=" << src1->src[1]->nb[3] << std::endl;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
std::cerr << std::endl << "Result:" << std::endl;
ggml_vk_print_tensor_area(src1_clone, src1_clone->data, 5, 5, 0, 0);
std::cerr << std::endl;
std::cerr << std::endl << "Result:" << std::endl;
ggml_vk_print_tensor_area(src1_clone, src1_clone->data, 5, 5, 1, 0);
std::cerr << std::endl;
std::vector<const ggml_tensor *> done;
ggml_vk_print_graph_origin(src1_clone, done);
}
}
if (src2 != nullptr) {
src2_clone = ggml_dup_tensor(ggml_ctx, src2);
src2_size = ggml_nbytes(src2);
src2_buffer = malloc(src2_size);
src2_clone->data = src2_buffer;
if (ggml_backend_buffer_is_host(src2->buffer)) {
memcpy(src2_clone->data, src2->data, src2_size);
memcpy(src2_clone->nb, src2->nb, sizeof(size_t) * GGML_MAX_DIMS);
} else if (ggml_backend_buffer_is_vk(src2->buffer)) {
ggml_tensor_extra_gpu * extra = (ggml_tensor_extra_gpu *) src2->extra;
vk_buffer buffer_gpu = extra->buffer_gpu.lock();
uint64_t offset = extra->offset + src2->view_offs;
if (!ggml_is_contiguous(src2) && ggml_vk_dim01_contiguous(src2)) {
for (int i3 = 0; i3 < src2->ne[3]; i3++) {
for (int i2 = 0; i2 < src2->ne[2]; i2++) {
const int idx = i3*src2->ne[2] + i2;
ggml_vk_buffer_read(ctx, buffer_gpu, offset + idx * src2->nb[2], ((char *)src2_clone->data + idx * src2_clone->nb[2]), src2->ne[1] * src2->nb[1]);
}
}
src2_clone->nb[0] = src2->nb[0];
src2_clone->nb[1] = src2->nb[1];
for (int i = 2; i < GGML_MAX_DIMS; i++) {
src2_clone->nb[i] = src2_clone->nb[i - 1]*src2_clone->ne[i - 1];
}
} else {
if (offset + src2_size >= buffer_gpu->size) {
src2_size = buffer_gpu->size - offset;
}
ggml_vk_buffer_read(ctx, buffer_gpu, offset, src2_clone->data, src2_size);
memcpy(src2_clone->nb, src2->nb, sizeof(size_t) * GGML_MAX_DIMS);
}
} else {
GGML_ASSERT(false);
}
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
if (vk_output_tensor > 0 && vk_output_tensor == check_counter) {
ggml_vk_print_tensor(ctx, src2, "src2");
std::cerr << "TENSOR CHECK: " << ggml_op_name(src2_clone->op) << " (check " << check_counter << ")" << std::endl;
std::cerr << "src2_clone=" << tensor << " src2_clone->type: " << ggml_type_name(src2_clone->type) << " ne0=" << src2_clone->ne[0] << " nb0=" << src2_clone->nb[0] << " ne1=" << src2_clone->ne[1] << " nb1=" << src2_clone->nb[1] << " ne2=" << src2_clone->ne[2] << " nb2=" << src2_clone->nb[2] << " ne3=" << src2_clone->ne[3] << " nb3=" << src2_clone->nb[3] << std::endl;
if (src2->src[0] != nullptr) {
std::cerr << "src2->src[0]=" << src2->src[0] << " op=" << ggml_op_name(src2->src[0]->op) << " type=" << ggml_type_name(src2->src[0]->type) << " ne0=" << src2->src[0]->ne[0] << " nb0=" << src2->src[0]->nb[0] << " ne1=" << src2->src[0]->ne[1] << " nb1=" << src2->src[0]->nb[1] << " ne2=" << src2->src[0]->ne[2] << " nb2=" << src2->src[0]->nb[2] << " ne3=" << src2->src[0]->ne[3] << " nb3=" << src2->src[0]->nb[3] << std::endl;
}
if (src2->src[1] != nullptr) {
std::cerr << "src2->src[1]=" << src2->src[1] << " op=" << ggml_op_name(src2->src[1]->op) << " type=" << ggml_type_name(src2->src[1]->type) << " ne0=" << src2->src[1]->ne[0] << " nb0=" << src2->src[1]->nb[0] << " ne1=" << src2->src[1]->ne[1] << " nb1=" << src2->src[1]->nb[1] << " ne2=" << src2->src[1]->ne[2] << " nb2=" << src2->src[1]->nb[2] << " ne3=" << src2->src[1]->ne[3] << " nb3=" << src2->src[1]->nb[3] << std::endl;
}
std::cerr << std::endl << "Result:" << std::endl;
ggml_vk_print_tensor_area(src2_clone, src2_clone->data, 5, 5, 0, 0);
std::cerr << std::endl;
std::cerr << std::endl << "Result:" << std::endl;
ggml_vk_print_tensor_area(src2_clone, src2_clone->data, 5, 5, 1, 0);
std::cerr << std::endl;
std::vector<const ggml_tensor *> done;
ggml_vk_print_graph_origin(src2_clone, done);
}
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
if (tensor->op == GGML_OP_MUL_MAT) {
tensor_clone = ggml_mul_mat(ggml_ctx, src0_clone, src1_clone);
} else if (tensor->op == GGML_OP_MUL_MAT_ID) {
tensor_clone = ggml_mul_mat_id(ggml_ctx, src0_clone, src1_clone, src2_clone);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
} else if (tensor->op == GGML_OP_MUL) {
tensor_clone = ggml_mul(ggml_ctx, src0_clone, src1_clone);
} else if (tensor->op == GGML_OP_DIV) {
tensor_clone = ggml_div(ggml_ctx, src0_clone, src1_clone);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
} else if (tensor->op == GGML_OP_SCALE) {
tensor_clone = ggml_scale(ggml_ctx, src0_clone, ((float *)tensor->op_params)[0]);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
} else if (tensor->op == GGML_OP_SQR) {
tensor_clone = ggml_sqr(ggml_ctx, src0_clone);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
} else if (tensor->op == GGML_OP_CLAMP) {
tensor_clone = ggml_clamp(ggml_ctx, src0_clone, ((float *)tensor->op_params)[0], ((float *)tensor->op_params)[1]);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
} else if (tensor->op == GGML_OP_ADD) {
tensor_clone = ggml_add(ggml_ctx, src0_clone, src1_clone);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
} else if (tensor->op == GGML_OP_NORM) {
tensor_clone = ggml_norm(ggml_ctx, src0_clone, *(float *)tensor->op_params);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
} else if (tensor->op == GGML_OP_RMS_NORM) {
tensor_clone = ggml_rms_norm(ggml_ctx, src0_clone, *(float *)tensor->op_params);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
} else if (tensor->op == GGML_OP_SOFT_MAX) {
if (src1 != nullptr) {
tensor_clone = ggml_soft_max_ext(ggml_ctx, src0_clone, src1_clone, ((float *)tensor->op_params)[0], ((float *)tensor->op_params)[1]);
} else {
tensor_clone = ggml_soft_max(ggml_ctx, src0_clone);
}
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
} else if (tensor->op == GGML_OP_DIAG_MASK_INF) {
tensor_clone = ggml_diag_mask_inf(ggml_ctx, src0_clone, *(int *)tensor->op_params);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
} else if (tensor->op == GGML_OP_ROPE) {
const int n_dims = ((int32_t *) tensor->op_params)[1];
const int mode = ((int32_t *) tensor->op_params)[2];
//const int n_ctx_ggml = ((int32_t *) tensor->op_params)[3];
const int n_ctx_orig_ggml = ((int32_t *) tensor->op_params)[4];
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
float freq_base = ((float *) tensor->op_params)[5];
float freq_scale = ((float *) tensor->op_params)[6];
float ext_factor = ((float *) tensor->op_params)[7];
float attn_factor = ((float *) tensor->op_params)[8];
float beta_fast = ((float *) tensor->op_params)[9];
float beta_slow = ((float *) tensor->op_params)[10];
tensor_clone = ggml_rope_ext(ggml_ctx, src0_clone, src1_clone, src2_clone, n_dims, mode, n_ctx_orig_ggml, freq_base, freq_scale, ext_factor, attn_factor, beta_fast, beta_slow);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
} else if (tensor->op == GGML_OP_UNARY) {
switch (ggml_get_unary_op(tensor)) {
case GGML_UNARY_OP_SILU:
tensor_clone = ggml_silu(ggml_ctx, src0_clone);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
break;
case GGML_UNARY_OP_GELU:
tensor_clone = ggml_gelu(ggml_ctx, src0_clone);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
break;
case GGML_UNARY_OP_RELU:
tensor_clone = ggml_relu(ggml_ctx, src0_clone);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
break;
default:
std::cerr << "Missing vk_check_results OP: " << ggml_op_name(tensor->op) << std::endl;
GGML_ASSERT(false);
}
} else if (tensor->op == GGML_OP_CPY || tensor->op == GGML_OP_DUP) {
if (src1 == nullptr) {
tensor_clone = ggml_dup(ggml_ctx, src0_clone);
tensor_clone->type = tensor->type;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
} else {
tensor_clone = ggml_cpy(ggml_ctx, src0_clone, src1_clone);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
} else if (tensor->op == GGML_OP_CONT) {
tensor_clone = ggml_cont_4d(ggml_ctx, src0_clone, tensor->ne[0], tensor->ne[1], tensor->ne[2], tensor->ne[3]);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
} else if (tensor->op == GGML_OP_RESHAPE) {
tensor_clone = ggml_reshape_4d(ggml_ctx, src0_clone, tensor->ne[0], tensor->ne[1], tensor->ne[2], tensor->ne[3]);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
} else if (tensor->op == GGML_OP_VIEW) {
tensor_clone = ggml_view_4d(ggml_ctx, src0_clone, tensor->ne[0], tensor->ne[1], tensor->ne[2], tensor->ne[3], tensor->nb[1], tensor->nb[2], tensor->nb[3], ((int32_t *) tensor->op_params)[0]);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
} else if (tensor->op == GGML_OP_PERMUTE) {
int32_t * params = (int32_t *)tensor->op_params;
tensor_clone = ggml_permute(ggml_ctx, src0_clone, params[0], params[1], params[2], params[3]);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
} else if (tensor->op == GGML_OP_TRANSPOSE) {
tensor_clone = ggml_transpose(ggml_ctx, src0_clone);
} else if (tensor->op == GGML_OP_GET_ROWS) {
tensor_clone = ggml_get_rows(ggml_ctx, src0_clone, src1_clone);
} else if (tensor->op == GGML_OP_ARGSORT) {
tensor_clone = ggml_argsort(ggml_ctx, src0_clone, (ggml_sort_order) *(int *)tensor->op_params);
} else if (tensor->op == GGML_OP_SUM_ROWS) {
tensor_clone = ggml_sum_rows(ggml_ctx, src0_clone);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
} else {
std::cerr << "Missing vk_check_results OP: " << ggml_op_name(tensor->op) << std::endl;
GGML_ASSERT(false);
}
ggml_cgraph * cgraph = ggml_new_graph(ggml_ctx);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
ggml_build_forward_expand(cgraph, tensor_clone);
ggml_graph_compute_with_ctx(ggml_ctx, cgraph, 8);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
if (vk_output_tensor > 0 && vk_output_tensor == check_counter) {
ggml_vk_print_tensor(ctx, tensor_clone, "tensor_clone");
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
comp_size = ggml_nbytes(tensor_clone);
comp_result = malloc(comp_size);
memcpy(comp_result, tensor_clone->data, comp_size);
memcpy(comp_nb, tensor_clone->nb, sizeof(size_t) * GGML_MAX_DIMS);
if (src0 != nullptr) {
free(src0_buffer);
}
if (src1 != nullptr) {
free(src1_buffer);
}
ggml_free(ggml_ctx);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
static void ggml_vk_check_results_1(ggml_backend_vk_context * ctx, ggml_tensor * tensor) {
if (tensor->op == GGML_OP_TRANSPOSE) {
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
return;
}
if (!(vk_output_tensor > 0 && vk_output_tensor == check_counter) && check_counter <= vk_skip_checks) {
return;
}
VK_LOG_DEBUG("ggml_vk_check_results_1(" << tensor->name << ")");
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
ggml_tensor * src0 = tensor->src[0];
ggml_tensor * src1 = tensor->src[1];
ggml_tensor * src2 = tensor->src[2];
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
void * tensor_data = tensor->data;
if (ggml_backend_buffer_is_vk(tensor->buffer)) {
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
size_t tensor_size = ggml_nbytes(tensor);
tensor_data = malloc(tensor_size);
ggml_tensor_extra_gpu * extra = (ggml_tensor_extra_gpu *) tensor->extra;
vk_buffer buffer_gpu = extra->buffer_gpu.lock();
if (extra->offset + tensor->view_offs + tensor_size >= buffer_gpu->size) {
tensor_size = buffer_gpu->size - (extra->offset + tensor->view_offs);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
ggml_vk_buffer_read(ctx, buffer_gpu, extra->offset + tensor->view_offs, tensor_data, tensor_size);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
float first_error_result = -1.0f;
float first_error_correct = -1.0f;
std::array<int, 4> first_error = { -1, -1, -1, -1 };
double avg_err = 0.0;
size_t counter = 0;
for (int i3 = 0; i3 < tensor->ne[3]; i3++) {
for (int i2 = 0; i2 < tensor->ne[2]; i2++) {
for (int i1 = 0; i1 < tensor->ne[1]; i1++) {
for (int i0 = 0; i0 < tensor->ne[0]; i0++) {
const bool buffer_size_fit = i3*comp_nb[3] + i2*comp_nb[2] + i1*comp_nb[1] + i0*comp_nb[0] < comp_size;
float correct = 0.0f;
float result = 0.0f;
if (buffer_size_fit) {
if (tensor->type == GGML_TYPE_F32) {
correct = *(float *) ((char *) comp_result + i3*comp_nb[3] + i2*comp_nb[2] + i1*comp_nb[1] + i0*comp_nb[0]);
result = *(float *) ((char *) tensor_data + i3*tensor->nb[3] + i2*tensor->nb[2] + i1*tensor->nb[1] + i0*tensor->nb[0]);
} else if (tensor->type == GGML_TYPE_F16) {
correct = ggml_fp16_to_fp32(*(ggml_fp16_t *) ((char *) comp_result + i3*comp_nb[3] + i2*comp_nb[2] + i1*comp_nb[1] + i0*comp_nb[0]));
result = ggml_fp16_to_fp32(*(ggml_fp16_t *) ((char *) tensor_data + i3*tensor->nb[3] + i2*tensor->nb[2] + i1*tensor->nb[1] + i0*tensor->nb[0]));
} else if (tensor->type == GGML_TYPE_I32) {
correct = *(int32_t *) ((char *) comp_result + i3*comp_nb[3] + i2*comp_nb[2] + i1*comp_nb[1] + i0*comp_nb[0]);
result = *(int32_t *) ((char *) tensor_data + i3*tensor->nb[3] + i2*tensor->nb[2] + i1*tensor->nb[1] + i0*tensor->nb[0]);
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
} else {
std::cerr << "Results check not implemented for type " << ggml_type_name(tensor->type) << std::endl;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
} else {
std::cerr << "Missing debug code for type " << ggml_type_name(tensor->type) << std::endl;
GGML_ASSERT(false);
}
if ((std::isnan(correct) != std::isnan(result)) || (std::isinf(correct) != std::isinf(result)) || !buffer_size_fit) {
std::cerr << "ERROR: Invalid value in " << ggml_op_name(tensor->op) << " i3=" << i3 << " i2=" << i2 << " i1=" << i1 << " i0=" << i0 << " result=" << result << " correct=" << correct << " avg_err=" << (avg_err / counter) << std::endl;
std::cerr << "tensor=" << tensor << " tensor->name=" << tensor->name << " tensor->type: " << ggml_type_name(tensor->type) << " ne0=" << tensor->ne[0] << " nb0=" << tensor->nb[0] << " ne1=" << tensor->ne[1] << " nb1=" << tensor->nb[1] << " ne2=" << tensor->ne[2] << " nb2=" << tensor->nb[2] << " ne3=" << tensor->ne[3] << " nb3=" << tensor->nb[3] << " offset=" << tensor->view_offs << std::endl;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
if (src0 != nullptr) {
std::cerr << "src0=" << src0 << " src0->name=" << src0->name << " op=" << ggml_op_name(src0->op) << " type=" << ggml_type_name(src0->type) << " ne0=" << src0->ne[0] << " nb0=" << src0->nb[0] << " ne1=" << src0->ne[1] << " nb1=" << src0->nb[1] << " ne2=" << src0->ne[2] << " nb2=" << src0->nb[2] << " ne3=" << src0->ne[3] << " nb3=" << src0->nb[3] << " offset=" << src0->view_offs << std::endl;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
if (src1 != nullptr) {
std::cerr << "src1=" << src1 << " src1->name=" << src1->name << " op=" << ggml_op_name(src1->op) << " type=" << ggml_type_name(src1->type) << " ne0=" << src1->ne[0] << " nb0=" << src1->nb[0] << " ne1=" << src1->ne[1] << " nb1=" << src1->nb[1] << " ne2=" << src1->ne[2] << " nb2=" << src1->nb[2] << " ne3=" << src1->ne[3] << " nb3=" << src1->nb[3] << " offset=" << src1->view_offs << std::endl;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
if (src2 != nullptr) {
std::cerr << "src2=" << src2 << " src2->name=" << src2->name << " op=" << ggml_op_name(src2->op) << " type=" << ggml_type_name(src2->type) << " ne0=" << src2->ne[0] << " nb0=" << src2->nb[0] << " ne1=" << src2->ne[1] << " nb1=" << src2->nb[1] << " ne2=" << src2->ne[2] << " nb2=" << src2->nb[2] << " ne3=" << src2->ne[3] << " nb3=" << src2->nb[3] << " offset=" << src2->view_offs << std::endl;
}
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
std::cerr << "First error: result=" << first_error_result << " correct=" << first_error_correct << " i3=" << first_error[3] << " i2=" << first_error[2] << " i1=" << first_error[1] << " i0=" << first_error[0] << std::endl;
std::cerr << std::endl << "Result:" << std::endl;
ggml_vk_print_tensor_area(tensor, tensor_data, i0, i1, i2, i3);
std::cerr << std::endl << "Correct:" << std::endl;
ggml_vk_print_tensor_area(tensor, comp_result, i0, i1, i2, i3);
std::cerr << std::endl;
std::vector<const ggml_tensor *> done;
ggml_vk_print_graph_origin(tensor, done);
GGML_ASSERT(false);
}
if (first_error[0] == -1 && std::fabs(correct - result) > 0.1f) {
first_error[0] = i0;
first_error[1] = i1;
first_error[2] = i2;
first_error[3] = i3;
first_error_result = result;
first_error_correct = correct;
}
// Special case, value is infinite, avoid NaN result in avg_err
// NaN also appears in results, if both are nan error is 0
if (!std::isinf(correct) && !std::isinf(result) && !std::isnan(correct) && !std::isnan(result)) {
avg_err += std::fabs(correct - result);
}
counter++;
}
}
}
}
avg_err /= counter;
if (vk_output_tensor > 0 && vk_output_tensor == check_counter) {
std::cerr << "TENSOR CHECK: avg_err=" << avg_err << " in " << ggml_op_name(tensor->op) << " (check " << check_counter << ")" << std::endl;
std::cerr << "tensor=" << tensor << " tensor->name=" << tensor->name << " tensor->type: " << ggml_type_name(tensor->type) << " ne0=" << tensor->ne[0] << " nb0=" << tensor->nb[0] << " ne1=" << tensor->ne[1] << " nb1=" << tensor->nb[1] << " ne2=" << tensor->ne[2] << " nb2=" << tensor->nb[2] << " ne3=" << tensor->ne[3] << " nb3=" << tensor->nb[3] << " offset=" << tensor->view_offs << std::endl;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
if (src0 != nullptr) {
std::cerr << "src0=" << src0 << " op=" << ggml_op_name(src0->op) << " type=" << ggml_type_name(src0->type) << " ne0=" << src0->ne[0] << " nb0=" << src0->nb[0] << " ne1=" << src0->ne[1] << " nb1=" << src0->nb[1] << " ne2=" << src0->ne[2] << " nb2=" << src0->nb[2] << " ne3=" << src0->ne[3] << " nb3=" << src0->nb[3] << " offset=" << src0->view_offs << std::endl;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
if (src1 != nullptr) {
std::cerr << "src1=" << src1 << " op=" << ggml_op_name(src1->op) << " type=" << ggml_type_name(src1->type) << " ne0=" << src1->ne[0] << " nb0=" << src1->nb[0] << " ne1=" << src1->ne[1] << " nb1=" << src1->nb[1] << " ne2=" << src1->ne[2] << " nb2=" << src1->nb[2] << " ne3=" << src1->ne[3] << " nb3=" << src1->nb[3] << " offset=" << src1->view_offs << std::endl;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
if (src2 != nullptr) {
std::cerr << "src2=" << src2 << " op=" << ggml_op_name(src2->op) << " type=" << ggml_type_name(src2->type) << " ne0=" << src2->ne[0] << " nb0=" << src2->nb[0] << " ne1=" << src2->ne[1] << " nb1=" << src2->nb[1] << " ne2=" << src2->ne[2] << " nb2=" << src2->nb[2] << " ne3=" << src2->ne[3] << " nb3=" << src2->nb[3] << " offset=" << src2->view_offs << std::endl;
}
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
std::cerr << "First error: result=" << first_error_result << " correct=" << first_error_correct << " i3=" << first_error[3] << " i2=" << first_error[2] << " i1=" << first_error[1] << " i0=" << first_error[0] << std::endl;
std::cerr << std::endl << "Result:" << std::endl;
ggml_vk_print_tensor_area(tensor, tensor_data, 5, 5, 0, 0);
std::cerr << std::endl << "Correct:" << std::endl;
ggml_vk_print_tensor_area(tensor, comp_result, 5, 5, 0, 0);
std::cerr << std::endl;
std::cerr << std::endl << "Result:" << std::endl;
ggml_vk_print_tensor_area(tensor, tensor_data, 5, 5, 1, 0);
std::cerr << std::endl << "Correct:" << std::endl;
ggml_vk_print_tensor_area(tensor, comp_result, 5, 5, 1, 0);
std::cerr << std::endl;
std::vector<const ggml_tensor *> done;
ggml_vk_print_graph_origin(tensor, done);
}
if (avg_err > 0.05 || std::isnan(avg_err)) {
std::cerr << "ERROR: avg_err=" << avg_err << " in " << ggml_op_name(tensor->op) << " (check " << check_counter << ")" << std::endl;
std::cerr << "tensor=" << tensor << " tensor->name=" << tensor->name << " tensor->type: " << ggml_type_name(tensor->type) << " ne0=" << tensor->ne[0] << " nb0=" << tensor->nb[0] << " ne1=" << tensor->ne[1] << " nb1=" << tensor->nb[1] << " ne2=" << tensor->ne[2] << " nb2=" << tensor->nb[2] << " ne3=" << tensor->ne[3] << " nb3=" << tensor->nb[3] << " offset=" << tensor->view_offs << std::endl;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
if (src0 != nullptr) {
std::cerr << "src0=" << src0 << " op=" << ggml_op_name(src0->op) << " type=" << ggml_type_name(src0->type) << " ne0=" << src0->ne[0] << " nb0=" << src0->nb[0] << " ne1=" << src0->ne[1] << " nb1=" << src0->nb[1] << " ne2=" << src0->ne[2] << " nb2=" << src0->nb[2] << " ne3=" << src0->ne[3] << " nb3=" << src0->nb[3] << " offset=" << src0->view_offs << std::endl;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
if (src1 != nullptr) {
std::cerr << "src1=" << src1 << " op=" << ggml_op_name(src1->op) << " type=" << ggml_type_name(src1->type) << " ne0=" << src1->ne[0] << " nb0=" << src1->nb[0] << " ne1=" << src1->ne[1] << " nb1=" << src1->nb[1] << " ne2=" << src1->ne[2] << " nb2=" << src1->nb[2] << " ne3=" << src1->ne[3] << " nb3=" << src1->nb[3] << " offset=" << src1->view_offs << std::endl;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
if (src2 != nullptr) {
std::cerr << "src2=" << src2 << " op=" << ggml_op_name(src2->op) << " type=" << ggml_type_name(src2->type) << " ne0=" << src2->ne[0] << " nb0=" << src2->nb[0] << " ne1=" << src2->ne[1] << " nb1=" << src2->nb[1] << " ne2=" << src2->ne[2] << " nb2=" << src2->nb[2] << " ne3=" << src2->ne[3] << " nb3=" << src2->nb[3] << " offset=" << src2->view_offs << std::endl;
}
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
std::cerr << "First error: result=" << first_error_result << " correct=" << first_error_correct << " i3=" << first_error[3] << " i2=" << first_error[2] << " i1=" << first_error[1] << " i0=" << first_error[0] << std::endl;
std::cerr << std::endl << "Result:" << std::endl;
ggml_vk_print_tensor_area(tensor, tensor_data, first_error[0], first_error[1], first_error[2], first_error[3]);
std::cerr << std::endl << "Correct:" << std::endl;
ggml_vk_print_tensor_area(tensor, comp_result, first_error[0], first_error[1], first_error[2], first_error[3]);
std::cerr << std::endl;
std::vector<const ggml_tensor *> done;
ggml_vk_print_graph_origin(tensor, done);
GGML_ASSERT(false);
} else {
std::cerr << check_counter << " " << tensor->name << " op=" << ggml_op_name(tensor->op) << " avg_err=" << avg_err << std::endl;
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
}
free(comp_result);
comp_result = nullptr;
comp_size = 0;
if (ggml_backend_buffer_is_vk(tensor->buffer)) {
ggml : add Vulkan backend (#2059) * Vulkan loader code * Fix matmul kernel, continue implementation * Continue implementation * Vulkan memory management * Vulkan development * Matmul call * Add aligned malloc and free for VMA * Continue implementation * First matmul success * GEMM Kernel optimization * 1D Blocktiling * 2D Blocktiling * Write coalescing * Continue vulkan implementation and optimization * First FP16 attempt, disabled for now * Code abstraction, FP16 implementation, fix kernel, add FP16 to FP32 kernel * Enable device extensions properly, restore fp16 matmul op * Fix mulmat_f16 * Output FP32 in fp16 matmul shader * Fix f16_to_f32 kernel * dequant_q4_0 kernel * Add VMA library * Avoid requesting dedicated memory, VMA can decide that by itself * Add bounds checking to matmul kernels, improve implementation, fix command buffers not freed properly * add cmake commands * Add 2d write operation, profiling code * Fix 2d write * Fix queue selection for AMD RADV * Fix trailing whitespace in vk_mem_alloc.h * Add WIP warp tile mat mul shaders * Disable glslc optimization * Disable glslc optimization for CMake * Optimize warptile matmul shader, replace blocktile with it * Add split-k optimization for small matrix multiplication Use semaphores for synchronization instead of fences or waitidle Rework async write/read for synchronization * Fix validation errors, improve compatibility with AMD GPUs * Rework command buffer handling * Variable matmul kernel using specialization constants * Fix synchronization on AMD, add barriers for buffer ownership transfer, add debug flag and prints * Reuse semaphores * Handle stage flags during command buffer submission properly * Increase matmul test runs for consistent results * Fix F32 matmul * Add vectorized loading and zeropadding for matrix multiplication * Use pinned memory for f16 preprocessing * Don't force aligned matmul * Don't free before queue done * Replace VMA library with native Vulkan buffer management * Basic offloading support with mul_f32 and dmmv for q4_0 * Run glslc commands in parallel * Unroll loops in dmmv shader * Reduce usage of waitIdle * Reuse pinned allocation for f16 conversion * Handle devices with only a single queue * Fix trailing whitespace in CMakeLists.txt * Allow parallel execution of kernels, parallelize third and fourth dimension calls * Add fallback for devices only supporting one DescriptorSet per DescriptorPool * Move to graph function similar to CUDA implementation * Use F16 kernel for most things, replace q_f32 with mul_mat_q_f16 function * Add F32 dmmv shaders * Batch submissions * Add .spv to gitignore * Split off matrix vector multiplication for separate optimization * Use single command buffer for matrix vector multiplication ops * Reduce overhead of mul_f32 calls by using a single command buffer * Add submission batching to mul_f32 * Fix tests * Add missing barrier * Add further missing barrier * Add further ops * Replace vk::QueueFamilyIgnored with VK_QUEUE_FAMILY_IGNORED to support more Vulkan header versions * Remove unnecessary cblas link * Fix descriptor set pre-allocation assert * Add runtime shader compilation, start transferring shaders to this approach * Transfer remaining shaders to header and compile on runtime * Fix fp32 fallback if device doesn't support fp16, add force disable env var GGML_VULKAN_DISABLE_F16 * Add support for q4_1, q5_0, q5_1 and q8_0 * Remove unnecessary scalar layout extension * Parse graph early to pre-record command buffers * Add q6_k support * Add multi-submit for command buffers * Fix q6_k dequant shader for AMD * Fix q6_k for GPUs without fp16 support * Simplify q6_k fp16 fix * Minor fixes * Fix wg_denom of m-mulmat shaders * Add Python-based Vulkan shader generator * Replace shaderc dependency with precompiled shaders Fix python script to generate shaders * Clean up code * Fix shader generator script Windows compatibility Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> * Close file before deletion * Fix vulkan shader fp32 name * Add q2_k and q3_k support Add validation check to compare shader results to cpu results * Add q4_k support * Add q5_k support * Bake SPIR-V bytecode into the library instead of loading shaders from file * Switch to signal semaphores for flexibility Prepare broadcasting support for mul mat * Finish broadcasting mul mat support for GQA * Clean up unused functions Add repeat op * Add further ops, not yet enabled. Improve semaphore code * Reduce number of used semaphores by utilizing timelines more properly * Remove queue information * Reuse timeline semaphores, allow parallel operation with binary semaphores to work around nvidia driver limitations * Add Vulkan to llama-bench * Remove cblas dependency * Fix matmul k-split bug * Fix q4_k dmmv K_QUANTS_PER_ITERATION 1 shader * Add RMS Norm shader, rework op_f32 shader setup, fix matmul bug * Fix issues with float16 overflows in shaders * Fix issues with older Vulkan headers on Ubuntu 22.04 * Allow multi-op partial offloading by parsing the graph to preallocate enough between-op buffers * Implement further ops, rework op_f32 calls, fix bugs * Finish full offloading support, add last remaining ops, fix bugs, remove redundant code * Upload generated file ggml-vulkan-shaders.hpp, remove redundant shaders * Merge upstream changes, fix conflicts, adapt soft_max op * Fix Python and shader header format * Free model gpu buffers on exit * Use single queue per device to simplify code * Add matmul shader support for running multiple calculations in parallel * Switch from semaphore-synchronized multiple command buffers per op to single command buffer for multiple ops, whole graph if possible * Fix missing event cast * Replace uint64_t(-1) with UINT64_MAX, rename function for clarity * Fix warning about empty C function parameters * Fix compiler warnings * Properly implement Vulkan backend buffer handling * Fix oversized host staging buffers * Simplify barrier synchronization calls * Fix gcc warnings * Implement max_size for backend buffer types to limit the size of a single allocation * Use min of maxMemoryAllocationSize and maxBufferSize for device max allocation size * refactor multi buf * Disable unsupported ops to fix tests * Check for maintenance4 support before using it * Handle devices with only a single queue * Fix single queue logic * propagate buffer usage in multi buffers * Implement rope_neox op * Cleanup header and other files * Simplify gpu_extras by removing events and putting staging memcpys into contexts * Move queue into context Add not-yet-enabled async backend ops * Simplify context use, optimize matmul shader for warp size 64 (AMD GCN), fix split_k matmul shader optimization * Add get_max_size to SYCL backend. Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * llama : fix trailing whitespace --------- Co-authored-by: Henri Vasserman <henv@hot.ee> Co-authored-by: Concedo <39025047+LostRuins@users.noreply.github.com> Co-authored-by: slaren <slarengh@gmail.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>
2024-01-28 18:03:59 +01:00
free(tensor_data);
}
}
#endif