GitHub - Tudi/ASM-YUV-image-processing-for-video-encoding: SAD, SATD, DCT, PR written inYASM and using SSE2

Branches Tags

Name		Name	Last commit message	Last commit date
Latest commit History 2 Commits
dep_libCPU		dep_libCPU
ASM_functions.sln		ASM_functions.sln
ASM_functions.vcproj		ASM_functions.vcproj
ASM_functions_2k8.vcproj		ASM_functions_2k8.vcproj
ASM_functions_VS2k8.sln		ASM_functions_VS2k8.sln
DCT.asm		DCT.asm
DCT_4x4_main.cpp		DCT_4x4_main.cpp
DCT_test.cpp		DCT_test.cpp
DCT_x64.asm		DCT_x64.asm
FilterHorBS_123_main.cpp		FilterHorBS_123_main.cpp
FilterVertBS_123_main.cpp		FilterVertBS_123_main.cpp
Interpolate.asm		Interpolate.asm
Interpolate.cpp		Interpolate.cpp
Prediction_Reader_SSE3.asm		Prediction_Reader_SSE3.asm
Prediction_Reader_SSE3_x64.asm		Prediction_Reader_SSE3_x64.asm
README.txt		README.txt
SAD.cpp		SAD.cpp
SAD_SSE2.asm		SAD_SSE2.asm
SAD_SSE2_Yasm.asm		SAD_SSE2_Yasm.asm
SAD_SSE2_Yasm_x64.asm		SAD_SSE2_Yasm_x64.asm
SAD_SSE3.asm		SAD_SSE3.asm
SAD_SSE3_x64.asm		SAD_SSE3_x64.asm
SAD_lib.sln		SAD_lib.sln
SAD_lib.vcproj		SAD_lib.vcproj
SAD_lib_vs10.sln		SAD_lib_vs10.sln
SAD_lib_vs10.vcxproj		SAD_lib_vs10.vcxproj
SAD_lib_vs9.sln		SAD_lib_vs9.sln
SAD_lib_vs9.vcproj		SAD_lib_vs9.vcproj
SATD.cpp		SATD.cpp
SATD_16x8_main.cpp		SATD_16x8_main.cpp
SATD_8x16_main.cpp		SATD_8x16_main.cpp
SATD_SSE2.asm		SATD_SSE2.asm
SATD_SSE3.asm		SATD_SSE3.asm
SATD_SSE3_x64.asm		SATD_SSE3_x64.asm
SATD_SSE4.asm		SATD_SSE4.asm
SATD_SSE4_x64.asm		SATD_SSE4_x64.asm
Separation_SSE3.asm		Separation_SSE3.asm
Separation_SSE3_x64.asm		Separation_SSE3_x64.asm
StdAfx.h		StdAfx.h
Variance_test.cpp		Variance_test.cpp
interpolate_h_main.cpp		interpolate_h_main.cpp
interpolate_main.cpp		interpolate_main.cpp
lib_test.sln		lib_test.sln
lib_test.vcproj		lib_test.vcproj
lib_test_vs10.sln		lib_test_vs10.sln
lib_test_vs10.vcxproj		lib_test_vs10.vcxproj
lib_test_vs8.sln		lib_test_vs8.sln
lib_test_vs8.vcproj		lib_test_vs8.vcproj
main.cpp		main.cpp
main.h		main.h
main_libtest.cpp		main_libtest.cpp
quantize_main.cpp		quantize_main.cpp
read_avg_prediction16x16_main.cpp		read_avg_prediction16x16_main.cpp
read_diagonal_down_right_prediction_4x4_main.cpp		read_diagonal_down_right_prediction_4x4_main.cpp
read_luma_inter_pred_MxN_full_half_main.cpp		read_luma_inter_pred_MxN_full_half_main.cpp
read_plane_prediction16x16_main.cpp		read_plane_prediction16x16_main.cpp
read_vertical_right_prediction_4x4_main.cpp		read_vertical_right_prediction_4x4_main.cpp
test_SAD.cpp		test_SAD.cpp
test_SATD.cpp		test_SATD.cpp
test_SSD.cpp		test_SSD.cpp
test_Sum.cpp		test_Sum.cpp
test_cached_SAD.cpp		test_cached_SAD.cpp
test_separate.cpp		test_separate.cpp
vsyasm-1.2.0-win64.zip		vsyasm-1.2.0-win64.zip
yasm.rules		yasm.rules

Repository files navigation

Project was created to avoid visual studio intrinsic code instrumentation and optimization
Yasm can compile ASM as is to binary code. This is required when cache level or Hardware specific optimizations are used.
Using function pointers these functions gave about 16% increased speed when integrated to H264 encoder project

In case your project does not use function pointers you should benchamrk the "call"" overhead versus code inlining.