# HG changeset patch
# User Dnyaneshwar Gorade <[email protected]>
# Date 1381828023 -19800
# Tue Oct 15 14:37:03 2013 +0530
# Node ID d5097713f8c141547afc3f5813de4f064d8b1926
# Parent 13883d2d305709c262650bf4da564e3398456be3
pixel-sse41.cpp: Modified sse_ss16 function using faster macro
PROCESS_SSE_SS4x1.
diff -r 13883d2d3057 -r d5097713f8c1 source/common/vec/pixel-sse41.cpp
--- a/source/common/vec/pixel-sse41.cpp Tue Oct 15 14:34:25 2013 +0530
+++ b/source/common/vec/pixel-sse41.cpp Tue Oct 15 14:37:03 2013 +0530
@@ -5415,7 +5415,7 @@
{
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--)
{
@@ -5423,8 +5423,10 @@
{
PROCESS_SSE_SS4x1(i);
- 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
