Merge branch 'Tencent:master' into gemm-quantize-x86

Tencent · Oct 29, 2024 · 952bedd · 952bedd
2 parents 9c85f62 + c32442a
commit 952bedd
Show file tree

Hide file tree

Showing 2 changed files with 9 additions and 0 deletions.
diff --git a/CMakeLists.txt b/CMakeLists.txt
@@ -517,6 +517,8 @@ else()
 
         unset(CMAKE_REQUIRED_FLAGS)
     elseif(CMAKE_CXX_COMPILER_ID MATCHES "Clang" AND CMAKE_CXX_SIMULATE_ID MATCHES "MSVC" AND CMAKE_CXX_COMPILER_FRONTEND_VARIANT MATCHES "MSVC")
+        check_cxx_compiler_flag("-mrecip=none" NCNN_COMPILER_SUPPORT_X86_RECIP_NONE)
+
         check_cxx_compiler_flag("/arch:AVX" NCNN_COMPILER_SUPPORT_X86_AVX)
 
         set(CMAKE_REQUIRED_FLAGS "/arch:AVX -mfma -mf16c")
@@ -543,6 +545,8 @@ else()
 
         unset(CMAKE_REQUIRED_FLAGS)
     else()
+        check_cxx_compiler_flag("-mrecip=none" NCNN_COMPILER_SUPPORT_X86_RECIP_NONE)
+
         check_cxx_compiler_flag("-mavx" NCNN_COMPILER_SUPPORT_X86_AVX)
 
         set(CMAKE_REQUIRED_FLAGS "-mfma -mf16c")

diff --git a/src/CMakeLists.txt b/src/CMakeLists.txt
@@ -403,6 +403,11 @@ if(NCNN_TARGET_ARCH STREQUAL "x86")
                 target_compile_options(ncnn PRIVATE -msimd128)
             endif()
         endif()
+
+        if(NCNN_COMPILER_SUPPORT_X86_RECIP_NONE)
+            # recip optimization causes precision loss
+            target_compile_options(ncnn PRIVATE -mrecip=none)
+        endif()
     endif()
 
     if(NOT NCNN_RUNTIME_CPU AND NCNN_AVX512)