FastConvertYUVToARGBRow_SSSE3 use 2 pack and then 2 stores, which works better on Core2

BUG=none
TEST=none
Review URL: http://webrtc-codereview.appspot.com/323007

git-svn-id: http://libyuv.googlecode.com/svn/trunk@103 16f28f9a-4ce2-e073-06de-1de4eb20be90

README.chromium

 Name: libyuv
 URL: http://code.google.com/p/libyuv/
-Version: 102
+Version: 103
 License: BSD
 License File: LICENSE
 

source/row_posix.cc

@@ -409,8 +409,8 @@ void OMITFP FastConvertYUVToARGBRow_SSSE3(const uint8* y_buf,  // rdi
     "punpcklbw   %%xmm5,%%xmm2                 \n"
     "movdqa      %%xmm0,%%xmm1                 \n"
     "punpcklwd   %%xmm2,%%xmm0                 \n"
-    "movdqa      %%xmm0,(%3)                   \n"
     "punpckhwd   %%xmm2,%%xmm1                 \n"
+    "movdqa      %%xmm0,(%3)                   \n"
     "movdqa      %%xmm1,0x10(%3)               \n"
     "lea         0x20(%3),%3                   \n"
     "sub         $0x8,%4                       \n"

source/row_win.cc

@@ -614,8 +614,8 @@ void FastConvertYUVToARGBRow_SSSE3(const uint8* y_buf,
     punpcklbw  xmm2, xmm5           // RA
     movdqa     xmm1, xmm0
     punpcklwd  xmm0, xmm2           // BGRA first 4 pixels
-    movdqa     [edx], xmm0
     punpckhwd  xmm1, xmm2           // BGRA next 4 pixels
+    movdqa     [edx], xmm0
     movdqa     [edx + 16], xmm1
     lea        edx,  [edx + 32]
 
@@ -803,10 +803,10 @@ void FastConvertYToARGBRow_SSE2(const uint8* y_buf,
     punpcklbw  xmm0, xmm0           // GG
     movdqa     xmm1, xmm0
     punpcklwd  xmm0, xmm0           // BGRA first 4 pixels
-    por        xmm0, xmm5
-    movdqa     [edx], xmm0
     punpckhwd  xmm1, xmm1           // BGRA next 4 pixels
+    por        xmm0, xmm5
     por        xmm1, xmm5
+    movdqa     [edx], xmm0
     movdqa     [edx + 16], xmm1
     lea        edx,  [edx + 32]