the real function name is sa8d_8x8_avx2 whereas the common function name is sa8d_8x8_avx2_avx2, that's why we got proper call. both are different.
On Tue, Dec 1, 2015 at 9:08 PM, chen <chenm...@163.com> wrote: > > > At 2015-12-01 18:58:43,dnyanesh...@multicorewareinc.com wrote: > ># HG changeset patch > ># User Dnyaneshwar G <dnyanesh...@multicorewareinc.com> > ># Date 1448962785 -19800 > ># Tue Dec 01 15:09:45 2015 +0530 > ># Node ID f8b0ce4e9f4092a38d8095961825e734a34f112e > ># Parent e2e507ffe752d6c193a219b242c433bdc55f39f7 > >asm: SA8D avx2 asm code for main12 > > > >sa8d[ 8x8] 4.70x 564.58 2652.82 > >sa8d[ 8x16] 4.00x 1358.06 5429.52 > >sa8d[16x16] 5.57x 2013.70 11212.47 > >sa8d[16x32] 3.90x 5610.47 21883.35 > >sa8d[32x32] 5.36x 8274.18 44361.61 > >sa8d[32x64] 3.86x 23024.04 88901.80 > >sa8d[64x64] 4.35x 45509.79 198165.11 > > > >diff -r e2e507ffe752 -r f8b0ce4e9f40 source/common/x86/asm-primitives.cpp > >--- a/source/common/x86/asm-primitives.cpp Mon Nov 30 11:23:38 2015 +0530 > >+++ b/source/common/x86/asm-primitives.cpp Tue Dec 01 15:09:45 2015 +0530 > >@@ -1313,6 +1313,9 @@ > > } > > if (cpuMask & X265_CPU_AVX2) > > { > >+#if X265_DEPTH == 12 > >+ ASSIGN_SA8D(avx2); > >+#endif > > p.cu[BLOCK_4x4].intra_filter = PFX(intra_filter_4x4_avx2); > > > > // TODO: the planecopy_sp is really planecopy_SC now, must be fix it > >diff -r e2e507ffe752 -r f8b0ce4e9f40 source/common/x86/pixel-a.asm > >--- a/source/common/x86/pixel-a.asm Mon Nov 30 11:23:38 2015 +0530 > >+++ b/source/common/x86/pixel-a.asm Tue Dec 01 15:09:45 2015 +0530 > >@@ -6499,6 +6499,1357 @@ > > %endif ; !ARCH_X86_64 > > %endmacro ; SA8D > > > >+ > >+%if ARCH_X86_64 == 1 && BIT_DEPTH == 12 > >+INIT_YMM avx2 > >+cglobal sa8d_8x8_avx2 > the really function name is sa8d_8x8_avx2_avx2, we are lucky, below call use > correct name > > > _______________________________________________ > x265-devel mailing list > x265-devel@videolan.org > https://mailman.videolan.org/listinfo/x265-devel > >
_______________________________________________ x265-devel mailing list x265-devel@videolan.org https://mailman.videolan.org/listinfo/x265-devel