On Tue, Oct 15, 2013 at 3:52 AM, <[email protected]> wrote:
> # HG changeset patch > # User Dnyaneshwar Gorade <[email protected]> > # Date 1381826441 -19800 > # Tue Oct 15 14:10:41 2013 +0530 > # Node ID e00905521f015292f9ccc1ae83b1476c3c433310 > # Parent 1019dc55ec1b7f6b4ba7d15126d96167998c4031 > pixel-sse41.cpp: Modified sse_ss8 function using faster macro > PROCESS_SSE_SS4x1. > these all need to be packaged together in one patch with the change in the macro itself, otherwise the build is broken for a stretch of commits, which is always bad when you're trying to isolate a change that introduced a bug > > diff -r 1019dc55ec1b -r e00905521f01 source/common/vec/pixel-sse41.cpp > --- a/source/common/vec/pixel-sse41.cpp Tue Oct 15 14:07:36 2013 +0530 > +++ b/source/common/vec/pixel-sse41.cpp Tue Oct 15 14:10:41 2013 +0530 > @@ -5361,14 +5361,16 @@ > { > int rows = ly; > __m128i sum = _mm_setzero_si128(); > - __m128i m1, n1, diff, sign1, sign2, tmp1, tmp2; > + __m128i m1, n1, diff, sign, tmp1, tmp2; > > for (; rows != 0; rows--) > { > PROCESS_SSE_SS4x1(0); > > - m1 = _mm_unpackhi_epi16(m1, sign1); > - n1 = _mm_unpackhi_epi16(n1, sign2); > + sign = _mm_srai_epi16(m1, 15); > + m1 = _mm_unpackhi_epi16(m1, sign); > + sign = _mm_srai_epi16(n1, 15); > + n1 = _mm_unpackhi_epi16(n1, sign); > diff = _mm_sub_epi32(m1, n1); > diff = _mm_mullo_epi32(diff, diff); > sum = _mm_add_epi32(sum, diff); > _______________________________________________ > x265-devel mailing list > [email protected] > https://mailman.videolan.org/listinfo/x265-devel > -- Steve Borho
_______________________________________________ x265-devel mailing list [email protected] https://mailman.videolan.org/listinfo/x265-devel
