commit 014fa4529802ec89d1f3239722955c1493b688db
author Jingning Han <jingning@google.com> Tue Mar 31 10:08:29 2015 -0700
committer Gerrit Code Review <gerrit@gerrit.golo.chromium.org> Tue Mar 31 10:21:52 2015 -0700
tree 2c22759f2d9dca324002c1edb099bc5d30522c21
parent ebe1be9186b4f20e8d30c4bbad79ba78403de0d2

Use aligned copy in 8x8 Hadamard transform SSE2

This reduces the 8x8 Hadamard transform cycles by 20%.

Change-Id: If34c5e02f3afa42244c6efabe121f7cf5d2df41b

vp9/encoder/x86/vp9_avg_intrin_sse2.c

diff --git a/vp9/encoder/x86/vp9_avg_intrin_sse2.c b/vp9/encoder/x86/vp9_avg_intrin_sse2.c
index f2c7c64..ecd6ce9 100644
--- a/vp9/encoder/x86/vp9_avg_intrin_sse2.c
+++ b/vp9/encoder/x86/vp9_avg_intrin_sse2.c
@@ -148,21 +148,21 @@
   hadamard_col8_sse2(src, 0);
   hadamard_col8_sse2(src, 1);
 
-  _mm_storeu_si128((__m128i *)coeff, src[0]);
+  _mm_store_si128((__m128i *)coeff, src[0]);
   coeff += 8;
-  _mm_storeu_si128((__m128i *)coeff, src[1]);
+  _mm_store_si128((__m128i *)coeff, src[1]);
   coeff += 8;
-  _mm_storeu_si128((__m128i *)coeff, src[2]);
+  _mm_store_si128((__m128i *)coeff, src[2]);
   coeff += 8;
-  _mm_storeu_si128((__m128i *)coeff, src[3]);
+  _mm_store_si128((__m128i *)coeff, src[3]);
   coeff += 8;
-  _mm_storeu_si128((__m128i *)coeff, src[4]);
+  _mm_store_si128((__m128i *)coeff, src[4]);
   coeff += 8;
-  _mm_storeu_si128((__m128i *)coeff, src[5]);
+  _mm_store_si128((__m128i *)coeff, src[5]);
   coeff += 8;
-  _mm_storeu_si128((__m128i *)coeff, src[6]);
+  _mm_store_si128((__m128i *)coeff, src[6]);
   coeff += 8;
-  _mm_storeu_si128((__m128i *)coeff, src[7]);
+  _mm_store_si128((__m128i *)coeff, src[7]);
 }
 
 void vp9_hadamard_16x16_sse2(int16_t const *src_diff, int src_stride,