saymrwulf/onnxruntime
1
0
Fork 0
mirror of https://github.com/saymrwulf/onnxruntime.git synced 2026-07-10 17:37:14 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

Build new AVX file with /ARCH:AVX (#4442)

Browse source 
Build new file with /ARCH:AVX on Windows to ensure correct vzeroupper behavior.
This commit is contained in:
Tracy Sharpe 2020-07-07 12:00:12 -07:00 committed by GitHub
parent e62686c36e
commit aa06d308a6
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
1 changed files with 8 additions and 4 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

12
cmake/onnxruntime_mlas.cmake
View file

@ -45,14 +45,20 @@ if(MSVC)
${ONNXRUNTIME_ROOT}/core/mlas/lib/arm/sgemmc.cpp
)
elseif(onnxruntime_target_platform STREQUAL "x64")
enable_language(ASM_MASM)
set(mlas_platform_srcs_avx
${ONNXRUNTIME_ROOT}/core/mlas/lib/intrinsics/avx/min_max_elements.cpp
)
set_source_files_properties(${mlas_platform_srcs_avx} PROPERTIES COMPILE_FLAGS "/arch:AVX")
set(mlas_platform_srcs_avx2
${ONNXRUNTIME_ROOT}/core/mlas/lib/intrinsics/avx2/qladd_avx2.cpp
)
set_source_files_properties(${mlas_platform_srcs_avx2} PROPERTIES COMPILE_FLAGS "/arch:AVX2")
enable_language(ASM_MASM)
set(mlas_platform_srcs
${mlas_platform_srcs_avx}
${mlas_platform_srcs_avx2}
${ONNXRUNTIME_ROOT}/core/mlas/lib/amd64/QgemmU8S8KernelAvx2.asm
${ONNXRUNTIME_ROOT}/core/mlas/lib/amd64/QgemvU8S8KernelAvx2.asm
@ -86,7 +92,6 @@ if(MSVC)
${ONNXRUNTIME_ROOT}/core/mlas/lib/amd64/LogisticKernelFma3.asm
${ONNXRUNTIME_ROOT}/core/mlas/lib/amd64/TanhKernelFma3.asm
${ONNXRUNTIME_ROOT}/core/mlas/lib/amd64/ErfKernelFma3.asm
${ONNXRUNTIME_ROOT}/core/mlas/lib/intrinsics/avx/min_max_elements.cpp
)
else()
enable_language(ASM_MASM)
@ -177,7 +182,6 @@ else()
# Forward the flags for the minimum target platform version from the C
# compiler to the assembler. This works around CMakeASMCompiler.cmake.in
# not including the logic to set this flag for the assembler.
set(CMAKE_ASM${ASM_DIALECT}_OSX_DEPLOYMENT_TARGET_FLAG "${CMAKE_C_OSX_DEPLOYMENT_TARGET_FLAG}")
# The LLVM assembler does not support the .arch directive to enable instruction