vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

arm intrinsics detection for msvc

Browse source
This commit is contained in:
Michael Podvitskiy 2024-02-07 22:12:47 +01:00
parent aa7ab99be2
commit 1d618a5efb
1 changed files with 13 additions and 3 deletions
Download patch file Download diff file Expand all files Collapse all files

14
CMakeLists.txt
View file

@ -853,11 +853,21 @@ set(ARCH_FLAGS "")
if ((${CMAKE_SYSTEM_PROCESSOR} MATCHES "arm") OR (${CMAKE_SYSTEM_PROCESSOR} MATCHES "aarch64") OR ("${CMAKE_GENERATOR_PLATFORM_LWR}" MATCHES "arm64")) if ((${CMAKE_SYSTEM_PROCESSOR} MATCHES "arm") OR (${CMAKE_SYSTEM_PROCESSOR} MATCHES "aarch64") OR ("${CMAKE_GENERATOR_PLATFORM_LWR}" MATCHES "arm64"))
message(STATUS "ARM detected") message(STATUS "ARM detected")
if (MSVC) if (MSVC)
add_compile_definitions(__aarch64__) # MSVC defines _M_ARM64 instead
add_compile_definitions(__ARM_NEON) add_compile_definitions(__ARM_NEON)
add_compile_definitions(__ARM_FEATURE_FMA) add_compile_definitions(__ARM_FEATURE_FMA)
set(CMAKE_REQUIRED_FLAGS_PREV ${CMAKE_REQUIRED_FLAGS})
string(JOIN " " CMAKE_REQUIRED_FLAGS ${CMAKE_REQUIRED_FLAGS} "/arch:armv8.2")
check_cxx_source_compiles("#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)
add_compile_definitions(__ARM_FEATURE_DOTPROD) add_compile_definitions(__ARM_FEATURE_DOTPROD)
# add_compile_definitions(__ARM_FEATURE_FP16_VECTOR_ARITHMETIC) # MSVC doesn't support vdupq_n_f16, vld1q_f16, vst1q_f16 endif ()
add_compile_definitions(__aarch64__) # MSVC defines _M_ARM64 instead check_cxx_source_compiles("#include <arm_neon.h>\nint main() { float16_t _a; float16x8_t _s = vdupq_n_f16(_a); return 0; }" GGML_COMPILER_SUPPORT_FP16_VECTOR_ARITHMETIC)
if (GGML_COMPILER_SUPPORT_FP16_VECTOR_ARITHMETIC)
add_compile_definitions(__ARM_FEATURE_FP16_VECTOR_ARITHMETIC)
endif ()
set(CMAKE_REQUIRED_FLAGS ${CMAKE_REQUIRED_FLAGS_PREV})
else() else()
check_cxx_compiler_flag(-mfp16-format=ieee COMPILER_SUPPORTS_FP16_FORMAT_I3E) check_cxx_compiler_flag(-mfp16-format=ieee COMPILER_SUPPORTS_FP16_FORMAT_I3E)
if (NOT "${COMPILER_SUPPORTS_FP16_FORMAT_I3E}" STREQUAL "") if (NOT "${COMPILER_SUPPORTS_FP16_FORMAT_I3E}" STREQUAL "")