[x265] [PATCH] pixel: fix the hash mismatch due to convert32to16_shr

[chen]

My asm patch depend on this patch, please submit after this patch verifyed.


在 2013-10-17 16:25:27，"Steve Borho" <steve at borho.org> 写道：






On Thu, Oct 17, 2013 at 2:43 AM, <yuvaraj at multicorewareinc.com> wrote:
# HG changeset patch
# User Yuvaraj Venkatesh <yuvaraj at multicorewareinc.com>
# Date 1381995774 -19800
#      Thu Oct 17 13:12:54 2013 +0530
# Node ID 6f7ce9050675df305f39c16f7aa6e43767eecb0e
# Parent  1d6b3626f1b3d4b8f8f181e66eaaa649f90bac02
pixel: fix the hash mismatch due to convert32to16_shr



I'm holding this patch hostage until there's a test function for these convert primitives :)
 
diff -r 1d6b3626f1b3 -r 6f7ce9050675 source/common/vec/pixel-sse3.cpp
--- a/source/common/vec/pixel-sse3.cpp  Wed Oct 16 20:38:44 2013 -0500
+++ b/source/common/vec/pixel-sse3.cpp  Thu Oct 17 13:12:54 2013 +0530
@@ -44,7 +44,7 @@
         im32 = _mm_loadu_si128((__m128i const*)org);
         im32 = _mm_sra_epi32(_mm_add_epi32(im32, round), _mm_cvtsi32_si128(shift));
         im16 = _mm_packs_epi32(im32, im32);
-        _mm_storeu_si128((__m128i*)dst, im16);
+        _mm_storel_epi64((__m128i*)dst, im16);

         org += 4;
         dst += 4;
@@ -636,7 +636,7 @@
 namespace x265 {
 void Setup_Vec_PixelPrimitives_sse3(EncoderPrimitives &p)
 {
-    //p.cvt32to16_shr = convert32to16_shr;
+    p.cvt32to16_shr = convert32to16_shr;
     p.cvt16to32_shl = convert16to32_shl;
     p.cvt16to16_shl = convert16to16_shl;

_______________________________________________
x265-devel mailing list
x265-devel at videolan.org
https://mailman.videolan.org/listinfo/x265-devel






--
Steve Borho
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.videolan.org/pipermail/x265-devel/attachments/20131017/d88da9ea/attachment.html>