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

Data alignment fix for SSIM.

Browse Source 
WebRtc_UWord64[2] wasn't always aligned to 128 bytes, which
is necessary for _mm_store_si128.  By declaring the 
variable as __m128i it will always be 128 bytes aligned.

Incorrect include files.

__m128i is defined in emmintrin.h for visual studio.  Extra include on mac and linux is not a problem.
Review URL: http://webrtc-codereview.appspot.com/239013

git-svn-id: http://webrtc.googlecode.com/svn/trunk@816 4adac7df-926f-26a2-2b94-8c16560cd09d
This commit is contained in:
frkoenig@google.com 2011-10-26 00:07:32 +00:00
parent 78c767f9ba
commit fc9bcef8c7
1 changed files with 22 additions and 10 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

32
src/modules/video_coding/main/test/video_metrics.cc
View File

@ -164,6 +164,7 @@ Ssim8x8C(WebRtc_UWord8 *s, WebRtc_Word32 sp,
}
#if defined(WEBRTC_USE_SSE2)
#include <emmintrin.h>
#include <xmmintrin.h>
static double
Ssim8x8Sse2(WebRtc_UWord8 *s, WebRtc_Word32 sp,
@ -200,28 +201,39 @@ Ssim8x8Sse2(WebRtc_UWord8 *s, WebRtc_Word32 sp,
const __m128i sum_r_32 = _mm_add_epi32(_mm_unpackhi_epi16(sum_r_16, z),
_mm_unpacklo_epi16(sum_r_16, z));
WebRtc_UWord64 sum_s_64[2];
WebRtc_UWord64 sum_r_64[2];
WebRtc_UWord64 sum_sq_s_64[2];
WebRtc_UWord64 sum_sq_r_64[2];
WebRtc_UWord64 sum_sxr_64[2];
__m128i sum_s_128;
__m128i sum_r_128;
__m128i sum_sq_s_128;
__m128i sum_sq_r_128;
__m128i sum_sxr_128;
_mm_store_si128 ((__m128i*)sum_s_64,
_mm_store_si128 (&sum_s_128,
_mm_add_epi64(_mm_unpackhi_epi32(sum_s_32, z),
_mm_unpacklo_epi32(sum_s_32, z)));
_mm_store_si128 ((__m128i*)sum_r_64,
_mm_store_si128 (&sum_r_128,
_mm_add_epi64(_mm_unpackhi_epi32(sum_r_32, z),
_mm_unpacklo_epi32(sum_r_32, z)));
_mm_store_si128 ((__m128i*)sum_sq_s_64,
_mm_store_si128 (&sum_sq_s_128,
_mm_add_epi64(_mm_unpackhi_epi32(sum_sq_s_32, z),
_mm_unpacklo_epi32(sum_sq_s_32, z)));
_mm_store_si128 ((__m128i*)sum_sq_r_64,
_mm_store_si128 (&sum_sq_r_128,
_mm_add_epi64(_mm_unpackhi_epi32(sum_sq_r_32, z),
_mm_unpacklo_epi32(sum_sq_r_32, z)));
_mm_store_si128 ((__m128i*)sum_sxr_64,
_mm_store_si128 (&sum_sxr_128,
_mm_add_epi64(_mm_unpackhi_epi32(sum_sxr_32, z),
_mm_unpacklo_epi32(sum_sxr_32, z)));
const WebRtc_UWord64 *sum_s_64 =
reinterpret_cast<WebRtc_UWord64*>(&sum_s_128);
const WebRtc_UWord64 *sum_r_64 =
reinterpret_cast<WebRtc_UWord64*>(&sum_r_128);
const WebRtc_UWord64 *sum_sq_s_64 =
reinterpret_cast<WebRtc_UWord64*>(&sum_sq_s_128);
const WebRtc_UWord64 *sum_sq_r_64 =
reinterpret_cast<WebRtc_UWord64*>(&sum_sq_r_128);
const WebRtc_UWord64 *sum_sxr_64 =
reinterpret_cast<WebRtc_UWord64*>(&sum_sxr_128);
const WebRtc_UWord64 sum_s = sum_s_64[0] + sum_s_64[1];
const WebRtc_UWord64 sum_r = sum_r_64[0] + sum_r_64[1];
const WebRtc_UWord64 sum_sq_s = sum_sq_s_64[0] + sum_sq_s_64[1];