perf(*): change multiversioning target clones to x86-64-v3 and x86-64-v4

Signed-off-by: k4yt3x <[email protected]>
k4yt3x · Dec 30, 2024 · a72c094 · a72c094
1 parent 7665cd2
commit a72c094
Show file tree

Hide file tree

Showing 5 changed files with 12 additions and 12 deletions.
diff --git a/CMakeLists.txt b/CMakeLists.txt
@@ -1,4 +1,4 @@
-cmake_minimum_required(VERSION 3.10)
+cmake_minimum_required(VERSION 3.10 FATAL_ERROR)
 project(video2x VERSION 6.3.1 LANGUAGES CXX)
 
 # The FindBoost module is removed in CMake 3.30
@@ -346,10 +346,10 @@ target_include_directories(libvideo2x PRIVATE
 )
 
 # Set the compile options for the shared library
-if(MSVC)
-    target_compile_options(libvideo2x PRIVATE $<$<CONFIG:Debug>:/Zi /DDEBUG>)
-else()
-    target_compile_options(libvideo2x PRIVATE -fPIC $<$<CONFIG:Debug>:-g -DDEBUG>)
+if(CMAKE_CXX_COMPILER_ID STREQUAL "MSVC")
+    target_compile_options(libvideo2x PRIVATE $<$<CONFIG:Debug>:/Zi /Od /MDd>)
+elseif(CMAKE_CXX_COMPILER_ID MATCHES "GNU|Clang")
+    target_compile_options(libvideo2x PRIVATE -fPIC $<$<CONFIG:Debug>:-g -O0>)
 endif()
 
 # Define the paths to the shared libraries

diff --git a/src/avutils.cpp b/src/avutils.cpp
@@ -135,7 +135,7 @@ AVPixelFormat get_encoder_default_pix_fmt(const AVCodec *encoder, AVPixelFormat
     return best_pix_fmt;
 }
 
-[[gnu::target_clones("default", "avx2", "avx512f")]]
+[[gnu::target_clones("arch=x86-64-v4", "arch=x86-64-v3", "default")]]
 float get_frame_diff(AVFrame *frame1, AVFrame *frame2) {
     if (!frame1 || !frame2) {
         logger()->error("Invalid frame(s) provided for comparison");

diff --git a/src/conversions.cpp b/src/conversions.cpp
@@ -11,7 +11,7 @@ namespace video2x {
 namespace conversions {
 
 // Convert AVFrame format
-[[gnu::target_clones("default", "avx2", "avx512f")]]
+[[gnu::target_clones("arch=x86-64-v4", "arch=x86-64-v3", "default")]]
 AVFrame *convert_avframe_pix_fmt(AVFrame *src_frame, AVPixelFormat pix_fmt) {
     AVFrame *dst_frame = av_frame_alloc();
     if (dst_frame == nullptr) {
@@ -68,7 +68,7 @@ AVFrame *convert_avframe_pix_fmt(AVFrame *src_frame, AVPixelFormat pix_fmt) {
 }
 
 // Convert AVFrame to ncnn::Mat by copying the data
-[[gnu::target_clones("default", "avx2", "avx512f")]]
+[[gnu::target_clones("arch=x86-64-v4", "arch=x86-64-v3", "default")]]
 ncnn::Mat avframe_to_ncnn_mat(AVFrame *frame) {
     AVFrame *converted_frame = nullptr;
 
@@ -108,7 +108,7 @@ ncnn::Mat avframe_to_ncnn_mat(AVFrame *frame) {
 }
 
 // Convert ncnn::Mat to AVFrame with a specified pixel format (this part is unchanged)
-[[gnu::target_clones("default", "avx2", "avx512f")]]
+[[gnu::target_clones("arch=x86-64-v4", "arch=x86-64-v3", "default")]]
 AVFrame *ncnn_mat_to_avframe(const ncnn::Mat &mat, AVPixelFormat pix_fmt) {
     int ret;
 

diff --git a/src/encoder.cpp b/src/encoder.cpp
@@ -254,7 +254,7 @@ int Encoder::init(
     return 0;
 }
 
-[[gnu::target_clones("default", "avx2", "avx512f")]]
+[[gnu::target_clones("arch=x86-64-v4", "arch=x86-64-v3", "default")]]
 int Encoder::write_frame(AVFrame *frame, int64_t frame_idx) {
     AVFrame *converted_frame = nullptr;
     int ret;
@@ -326,7 +326,7 @@ int Encoder::write_frame(AVFrame *frame, int64_t frame_idx) {
     return 0;
 }
 
-[[gnu::target_clones("default", "avx2", "avx512f")]]
+[[gnu::target_clones("arch=x86-64-v4", "arch=x86-64-v3", "default")]]
 int Encoder::flush() {
     int ret;
     AVPacket *enc_pkt = av_packet_alloc();

diff --git a/src/libvideo2x.cpp b/src/libvideo2x.cpp
@@ -28,7 +28,7 @@ VideoProcessor::VideoProcessor(
       hw_device_type_(hw_device_type),
       benchmark_(benchmark) {}
 
-[[gnu::target_clones("default", "avx2", "avx512f")]]
+[[gnu::target_clones("arch=x86-64-v4", "arch=x86-64-v3", "default")]]
 int VideoProcessor::process(
     const std::filesystem::path in_fname,
     const std::filesystem::path out_fname