Mirrors/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggerganov/llama.cpp.git synced 2024-12-26 14:20:31 +01:00
Code Issues Packages Projects Releases Wiki Activity

ggml: fix arm build with gcc (#10895)

Browse Source 
Signed-off-by: Adrien Gallouët <angt@huggingface.co>
This commit is contained in:
Adrien Gallouët 2024-12-19 14:20:41 +01:00 committed by GitHub
parent 2fffc52b50
commit a3c33b1dce
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 21 additions and 8 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

29
ggml/src/ggml-cpu/CMakeLists.txt
View File

@ -88,32 +88,45 @@ function(ggml_add_cpu_backend_variant_impl tag_name)
endif()
if (GGML_NATIVE)
list(APPEND ARCH_FLAGS -mcpu=native)
set(CMAKE_REQUIRED_FLAGS_SAVE ${CMAKE_REQUIRED_FLAGS})
# -mcpu=native does not always enable all the features in some compilers,
# so we check for them manually and enable them if available
execute_process(
COMMAND ${CMAKE_C_COMPILER} -mcpu=native -E -v -
INPUT_FILE "/dev/null"
OUTPUT_QUIET
ERROR_VARIABLE ARM_MCPU
RESULT_VARIABLE ARM_MCPU_RESULT
)
if (NOT ARM_MCPU_RESULT)
string(REGEX MATCH "-mcpu=[^ ']+" ARM_MCPU_FLAG "${ARM_MCPU}")
endif()
if ("${ARM_MCPU_FLAG}" STREQUAL "")
set(ARM_MCPU_FLAG -mcpu=native)
message(STATUS "ARM -mcpu not found, -mcpu=native will be used")
endif()
set(CMAKE_REQUIRED_FLAGS_SAVE ${CMAKE_REQUIRED_FLAGS})
include(CheckCXXSourceRuns)
set(CMAKE_REQUIRED_FLAGS "${ARCH_FLAGS}+dotprod")
set(CMAKE_REQUIRED_FLAGS "${ARM_MCPU_FLAG}+dotprod")
check_cxx_source_runs(
"#include <arm_neon.h>\nint main() { int8x16_t _a, _b; int32x4_t _s = vdotq_s32(_s, _a, _b); return 0; }"
GGML_COMPILER_SUPPORT_DOTPROD)
if (GGML_COMPILER_SUPPORT_DOTPROD)
set(ARCH_FLAGS "${ARCH_FLAGS}+dotprod")
set(ARM_MCPU_FLAG_FIX "${ARM_MCPU_FLAG_FIX}+dotprod")
endif()
set(CMAKE_REQUIRED_FLAGS "${ARCH_FLAGS}+i8mm")
set(CMAKE_REQUIRED_FLAGS "${ARM_MCPU_FLAG}+i8mm")
check_cxx_source_runs(
"#include <arm_neon.h>\nint main() { int8x16_t _a, _b; int32x4_t _s = vmmlaq_s32(_s, _a, _b); return 0; }"
GGML_COMPILER_SUPPORT_I8MM)
if (GGML_COMPILER_SUPPORT_I8MM)
set(ARCH_FLAGS "${ARCH_FLAGS}+i8mm")
set(ARM_MCPU_FLAG_FIX "${ARM_MCPU_FLAG_FIX}+i8mm")
endif()
set(CMAKE_REQUIRED_FLAGS ${CMAKE_REQUIRED_FLAGS_SAVE})
list(APPEND ARCH_FLAGS "${ARM_MCPU_FLAG}${ARM_MCPU_FLAG_FIX}")
else()
if (GGML_CPU_ARM_ARCH)