Mirrors/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggerganov/llama.cpp.git synced 2025-02-04 23:52:32 +01:00
Code Issues Packages Projects Releases Wiki Activity

Restore hardcoded 32 in init_iq_shmem (AMD performance regression)

Browse Source
This commit is contained in:
Rémy O 2025-01-25 14:26:02 +01:00
parent 6ed3047e41
commit 3f7aa9dcb1
1 changed files with 20 additions and 10 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

20
ggml/src/ggml-vulkan/vulkan-shaders/types.comp
View File

@ -383,9 +383,11 @@ shared uvec2 iq2xxs_grid[256];
void init_iq_shmem() void init_iq_shmem()
{ {
// copy the table into shared memory and sync // copy the table into shared memory and sync
for (uint i = gl_LocalInvocationIndex.x; i < iq2xxs_grid.length(); i += gl_WorkGroupSize.x) { if (gl_LocalInvocationIndex.x < 32) {
for (uint i = gl_LocalInvocationIndex.x; i < iq2xxs_grid.length(); i += 32) {
iq2xxs_grid[i] = iq2xxs_grid_const[i]; iq2xxs_grid[i] = iq2xxs_grid_const[i];
} }
}
barrier(); barrier();
} }
@ -550,9 +552,11 @@ shared uvec2 iq2xs_grid[512];
void init_iq_shmem() void init_iq_shmem()
{ {
// copy the table into shared memory and sync // copy the table into shared memory and sync
for (uint i = gl_LocalInvocationIndex.x; i < iq2xs_grid.length(); i += gl_WorkGroupSize.x) { if (gl_LocalInvocationIndex.x < 32) {
for (uint i = gl_LocalInvocationIndex.x; i < iq2xs_grid.length(); i += 32) {
iq2xs_grid[i] = iq2xs_grid_const[i]; iq2xs_grid[i] = iq2xs_grid_const[i];
} }
}
barrier(); barrier();
} }
@ -839,9 +843,11 @@ shared uvec2 iq2s_grid[1024];
void init_iq_shmem() void init_iq_shmem()
{ {
// copy the table into shared memory and sync // copy the table into shared memory and sync
for (uint i = gl_LocalInvocationIndex.x; i < iq2s_grid.length(); i += gl_WorkGroupSize.x) { if (gl_LocalInvocationIndex.x < 32) {
for (uint i = gl_LocalInvocationIndex.x; i < iq2s_grid.length(); i += 32) {
iq2s_grid[i] = iq2s_grid_const[i]; iq2s_grid[i] = iq2s_grid_const[i];
} }
}
barrier(); barrier();
} }
@ -907,9 +913,11 @@ shared uint32_t iq3xxs_grid[256];
void init_iq_shmem() void init_iq_shmem()
{ {
// copy the table into shared memory and sync // copy the table into shared memory and sync
for (uint i = gl_LocalInvocationIndex.x; i < iq3xxs_grid.length(); i += gl_WorkGroupSize.x) { if (gl_LocalInvocationIndex.x < 32) {
for (uint i = gl_LocalInvocationIndex.x; i < iq3xxs_grid.length(); i += 32) {
iq3xxs_grid[i] = iq3xxs_grid_const[i]; iq3xxs_grid[i] = iq3xxs_grid_const[i];
} }
}
barrier(); barrier();
} }
@ -1014,9 +1022,11 @@ shared uint32_t iq3s_grid[512];
void init_iq_shmem() void init_iq_shmem()
{ {
// copy the table into shared memory and sync // copy the table into shared memory and sync
for (uint i = gl_LocalInvocationIndex.x; i < iq3s_grid.length(); i += gl_WorkGroupSize.x) { if (gl_LocalInvocationIndex.x < 32) {
for (uint i = gl_LocalInvocationIndex.x; i < iq3s_grid.length(); i += 32) {
iq3s_grid[i] = iq3s_grid_const[i]; iq3s_grid[i] = iq3s_grid_const[i];
} }
}
barrier(); barrier();
} }