generic-library/vpx
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

Use aligned copy in 8x8 Hadamard transform SSE2

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

Change-Id: If34c5e02f3afa42244c6efabe121f7cf5d2df41b
This commit is contained in:
Jingning Han 2015-03-31 10:08:29 -07:00 committed by Gerrit Code Review
parent ebe1be9186
commit 014fa45298
1 changed files with 8 additions and 8 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

16
vp9/encoder/x86/vp9_avg_intrin_sse2.c
View File

@ -148,21 +148,21 @@ void vp9_hadamard_8x8_sse2(int16_t const *src_diff, int src_stride,
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,