generic-library/openh264
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

[Processing] Add AVX2 VAA routines

Browse Source 
Process 8 lines at a time rather than 16 lines at a time because
this appears to give more reliable memory subsystem performance on
Haswell.

Speedup is > 2x as compared to SSE2 when not memory-bound on Haswell.
On my Haswell MBP, VAACalcSadSsdBgd is about ~3x faster when uncached,
which appears to be related to processing 8 lines at a time as opposed
to 16 lines at a time. The other routines are also faster as compared
to the SSE2 routines in this case but to a lesser extent.
This commit is contained in:
Sindre Aamås 2016-04-08 17:05:38 +02:00
parent eb9f56584f
commit 57fc3e9917
4 changed files with 1547 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

7
codec/processing/src/vaacalc/vaacalculation.cpp
View File

@ -64,6 +64,13 @@ void CVAACalculation::InitVaaFuncs (SVaaFuncs& sVaaFuncs, int32_t iCpuFlag) {
sVaaFuncs.pfVAACalcSadSsdBgd = VAACalcSadSsdBgd_sse2;
sVaaFuncs.pfVAACalcSadVar = VAACalcSadVar_sse2;
}
if (iCpuFlag & WELS_CPU_AVX2) {
sVaaFuncs.pfVAACalcSad = VAACalcSad_avx2;
sVaaFuncs.pfVAACalcSadBgd = VAACalcSadBgd_avx2;
sVaaFuncs.pfVAACalcSadSsd = VAACalcSadSsd_avx2;
sVaaFuncs.pfVAACalcSadSsdBgd = VAACalcSadSsdBgd_avx2;
sVaaFuncs.pfVAACalcSadVar = VAACalcSadVar_avx2;
}
#endif//X86_ASM
#ifdef HAVE_NEON
if ((iCpuFlag & WELS_CPU_NEON) == WELS_CPU_NEON) {

5
codec/processing/src/vaacalc/vaacalculation.h
View File

@ -104,6 +104,11 @@ VAACalcSadSsdBgdFunc VAACalcSadSsdBgd_sse2;
VAACalcSadFunc VAACalcSad_sse2;
VAACalcSadVarFunc VAACalcSadVar_sse2;
VAACalcSadSsdFunc VAACalcSadSsd_sse2;
VAACalcSadBgdFunc VAACalcSadBgd_avx2;
VAACalcSadSsdBgdFunc VAACalcSadSsdBgd_avx2;
VAACalcSadFunc VAACalcSad_avx2;
VAACalcSadVarFunc VAACalcSadVar_avx2;
VAACalcSadSsdFunc VAACalcSadSsd_avx2;
WELSVP_EXTERN_C_END
#endif

1529
codec/processing/src/x86/vaa.asm
View File

File diff suppressed because it is too large Load Diff

6
test/processing/ProcessUT_VaaCalc.cpp
View File

@ -828,6 +828,12 @@ GENERATE_VAACalcSadBgd_UT (VAACalcSadBgd_sse2, 1, WELS_CPU_SSE2)
GENERATE_VAACalcSadSsdBgd_UT (VAACalcSadSsdBgd_sse2, 1, WELS_CPU_SSE2)
GENERATE_VAACalcSadSsd_UT (VAACalcSadSsd_sse2, 1, WELS_CPU_SSE2)
GENERATE_VAACalcSadVar_UT (VAACalcSadVar_sse2, 1, WELS_CPU_SSE2)
GENERATE_VAACalcSad_UT (VAACalcSad_avx2, 1, WELS_CPU_AVX2)
GENERATE_VAACalcSadBgd_UT (VAACalcSadBgd_avx2, 1, WELS_CPU_AVX2)
GENERATE_VAACalcSadSsdBgd_UT (VAACalcSadSsdBgd_avx2, 1, WELS_CPU_AVX2)
GENERATE_VAACalcSadSsd_UT (VAACalcSadSsd_avx2, 1, WELS_CPU_AVX2)
GENERATE_VAACalcSadVar_UT (VAACalcSadVar_avx2, 1, WELS_CPU_AVX2)
#endif
#if defined(HAVE_NEON)