potok/opencv
1
0
Fork 0
mirror of https://github.com/opencv/opencv.git synced 2026-07-31 08:13:04 +04:00
Code Packages Releases Wiki Activity

DNN: add the Winograd fp16 support (#23654)

Browse Source 
* add Winograd FP16 implementation

* fixed dispatching of FP16 code paths in dnn; use dynamic dispatcher only when NEON_FP16 is enabled in the build and the feature is present in the host CPU at runtime

* fixed some warnings

* hopefully fixed winograd on x64 (and maybe other platforms)

---------

Co-authored-by: Vadim Pisarevsky <vadim.pisarevsky@gmail.com>
This commit is contained in:
zihaomu
2023-11-20 18:45:37 +08:00
committed by GitHub
parent a478757483
commit b913e73d04
22 changed files with 1262 additions and 556 deletions
Download Patch File Download Diff File Expand all files Collapse all files
cmake/OpenCVCompilerOptimizations.cmake
+1 -1
View File
@@ -49,7 +49,7 @@
set(CPU_ALL_OPTIMIZATIONS "SSE;SSE2;SSE3;SSSE3;SSE4_1;SSE4_2;POPCNT;AVX;FP16;AVX2;FMA3;AVX_512F")
list(APPEND CPU_ALL_OPTIMIZATIONS "AVX512_COMMON;AVX512_KNL;AVX512_KNM;AVX512_SKX;AVX512_CNL;AVX512_CLX;AVX512_ICL")
list(APPEND CPU_ALL_OPTIMIZATIONS NEON VFPV3 FP16 NEON_DOTPROD)
list(APPEND CPU_ALL_OPTIMIZATIONS NEON VFPV3 FP16 NEON_DOTPROD NEON_FP16 NEON_BF16)
list(APPEND CPU_ALL_OPTIMIZATIONS MSA)
list(APPEND CPU_ALL_OPTIMIZATIONS VSX VSX3)
list(APPEND CPU_ALL_OPTIMIZATIONS RVV)
cmake/checks/cpu_fp16.cpp
+5 -5
View File
@@ -15,12 +15,12 @@ int test()
#include "arm_neon.h"
int test()
{
const float src[] = { 0.0f, 0.0f, 0.0f, 0.0f };
short dst[8];
float32x4_t v_src = *(float32x4_t*)src;
const float src[] = { 0.0f, 1.0f, 2.0f, 3.0f };
short dst[4];
float32x4_t v_src = vld1q_f32(src);
float16x4_t v_dst = vcvt_f16_f32(v_src);
*(float16x4_t*)dst = v_dst;
return (int)dst[0];
vst1_f16((__fp16*)dst, v_dst);
return dst[0] + dst[1] + dst[2] + dst[3];
}
#else
#error "FP16 is not supported"