Load matrix with one vector and splat to 4 different ones.

BUG=none
TEST=none
R=ryanpetrie@google.com

Review URL: https://webrtc-codereview.appspot.com/3299004

git-svn-id: http://libyuv.googlecode.com/svn/trunk@838 16f28f9a-4ce2-e073-06de-1de4eb20be90

README.chromium

 Name: libyuv
 URL: http://code.google.com/p/libyuv/
-Version: 837
+Version: 838
 License: BSD
 License File: LICENSE
 

include/libyuv/version.h

@@ -11,6 +11,6 @@
 #ifndef INCLUDE_LIBYUV_VERSION_H_  // NOLINT
 #define INCLUDE_LIBYUV_VERSION_H_
 
-#define LIBYUV_VERSION 837
+#define LIBYUV_VERSION 838
 
 #endif  // INCLUDE_LIBYUV_VERSION_H_  NOLINT

source/row_posix.cc

@@ -4477,10 +4477,11 @@ void ARGBSepiaRow_SSSE3(uint8* dst_argb, int width) {
 void ARGBColorMatrixRow_SSSE3(const uint8* src_argb, uint8* dst_argb,
                               const int8* matrix_argb, int width) {
   asm volatile (
-    "pshufd    $0x00," MEMACCESS(3) ",%%xmm2   \n"
-    "pshufd    $0x55," MEMACCESS(3) ",%%xmm3   \n"
-    "pshufd    $0xaa," MEMACCESS(3) ",%%xmm4   \n"
-    "pshufd    $0xff," MEMACCESS(3) ",%%xmm5   \n"
+    "movdqu    " MEMACCESS(3) ",%%xmm5         \n"
+    "pshufd    $0x00,%%xmm5,%%xmm2             \n"
+    "pshufd    $0x55,%%xmm5,%%xmm3             \n"
+    "pshufd    $0xaa,%%xmm5,%%xmm4             \n"
+    "pshufd    $0xff,%%xmm5,%%xmm5             \n"
 
     // 8 pixel loop.
     ".p2align  4                               \n"

source/row_win.cc

@@ -5146,10 +5146,11 @@ void ARGBColorMatrixRow_SSSE3(const uint8* src_argb, uint8* dst_argb,
     mov        eax, [esp + 4]   /* src_argb */
     mov        edx, [esp + 8]   /* dst_argb */
     mov        ecx, [esp + 12]  /* matrix_argb */
-    pshufd     xmm2, [ecx], 0x00
-    pshufd     xmm3, [ecx], 0x55
-    pshufd     xmm4, [ecx], 0xaa
-    pshufd     xmm5, [ecx], 0xff
+    movdqu     xmm5, [ecx]
+    pshufd     xmm2, xmm5, 0x00
+    pshufd     xmm3, xmm5, 0x55
+    pshufd     xmm4, xmm5, 0xaa
+    pshufd     xmm5, xmm5, 0xff
     mov        ecx, [esp + 16]  /* width */
 
     align      4