Port some of the conversion routines to nacl

BUG=253
TEST=validator
R=nfullagar@google.com, ryanpetrie@google.com

Review URL: https://webrtc-codereview.appspot.com/1983004

git-svn-id: http://libyuv.googlecode.com/svn/trunk@748 16f28f9a-4ce2-e073-06de-1de4eb20be90

README.chromium

 Name: libyuv
 URL: http://code.google.com/p/libyuv/
-Version: 747
+Version: 748
 License: BSD
 License File: LICENSE
 

include/libyuv/row.h

@@ -44,15 +44,21 @@ extern "C" {
 #define HAS_ARGBBLENDROW_SSSE3
 #define HAS_ARGBCOLORMATRIXROW_SSSE3
 #define HAS_ARGBGRAYROW_SSSE3
+#define HAS_ARGBMIRRORROW_SSSE3
 #define HAS_ARGBMULTIPLYROW_SSE2
 #define HAS_ARGBQUANTIZEROW_SSE2
 #define HAS_ARGBSEPIAROW_SSSE3
 #define HAS_ARGBSHADEROW_SSE2
 #define HAS_ARGBSUBTRACTROW_SSE2
+#define HAS_COMPUTECUMULATIVESUMROW_SSE2
 
 // Conversions:
+#define HAS_ARGBTOBAYERROW_SSSE3
+#define HAS_ARGBSHUFFLEROW_SSSE3
 #define HAS_FIXEDDIV_X86
-
+#define HAS_ARGBTOYROW_SSSE3
+#define HAS_ARGBTOYJROW_SSSE3
+#define HAS_I400TOARGBROW_SSE2
 #endif
 
 // The following are available on all x86 platforms except NaCL x64:
@@ -65,10 +71,8 @@ extern "C" {
 #define HAS_ABGRTOYROW_SSSE3
 #define HAS_ARGB1555TOARGBROW_SSE2
 #define HAS_ARGB4444TOARGBROW_SSE2
-#define HAS_ARGBSHUFFLEROW_SSSE3
 #define HAS_ARGBTOARGB1555ROW_SSE2
 #define HAS_ARGBTOARGB4444ROW_SSE2
-#define HAS_ARGBTOBAYERROW_SSSE3
 #define HAS_ARGBTORAWROW_SSSE3
 #define HAS_ARGBTORGB24ROW_SSSE3
 #define HAS_ARGBTORGB565ROW_SSE2
@@ -76,15 +80,12 @@ extern "C" {
 #define HAS_ARGBTOUV444ROW_SSSE3
 #define HAS_ARGBTOUVROW_SSSE3
 #define HAS_ARGBTOUVJROW_SSSE3
-#define HAS_ARGBTOYROW_SSSE3
-#define HAS_ARGBTOYJROW_SSSE3
 #define HAS_BGRATOUVROW_SSSE3
 #define HAS_BGRATOYROW_SSSE3
 #define HAS_COPYROW_SSE2
 #define HAS_COPYROW_X86
 #define HAS_COPYROW_ERMS
 #define HAS_HALFROW_SSE2
-#define HAS_I400TOARGBROW_SSE2
 #define HAS_I411TOARGBROW_SSSE3
 #define HAS_I422TOABGRROW_SSSE3
 #define HAS_I422TOARGB1555ROW_SSSE3
@@ -126,9 +127,7 @@ extern "C" {
 
 // Effects:
 #define HAS_ARGBAFFINEROW_SSE2
-#define HAS_ARGBMIRRORROW_SSSE3
 #define HAS_ARGBUNATTENUATEROW_SSE2
-#define HAS_COMPUTECUMULATIVESUMROW_SSE2
 #define HAS_CUMULATIVESUMTOAVERAGEROW_SSE2
 #define HAS_INTERPOLATEROW_SSE2
 #define HAS_INTERPOLATEROW_SSSE3

include/libyuv/version.h

@@ -11,6 +11,6 @@
 #ifndef INCLUDE_LIBYUV_VERSION_H_  // NOLINT
 #define INCLUDE_LIBYUV_VERSION_H_
 
-#define LIBYUV_VERSION 747
+#define LIBYUV_VERSION 748
 
 #endif  // INCLUDE_LIBYUV_VERSION_H_  NOLINT

source/row_win.cc

@@ -3322,12 +3322,13 @@ void ARGBMirrorRow_SSSE3(const uint8* src, uint8* dst, int width) {
     mov       eax, [esp + 4]   // src
     mov       edx, [esp + 8]   // dst
     mov       ecx, [esp + 12]  // width
+    lea       eax, [eax - 16 + ecx * 4]  // last 4 pixels.
     movdqa    xmm5, kARGBShuffleMirror
-    lea       eax, [eax - 16]
 
     align      16
  convertloop:
-    movdqa    xmm0, [eax + ecx * 4]
+    movdqa    xmm0, [eax]
+    lea       eax, [eax - 16]
     pshufb    xmm0, xmm5
     sub       ecx, 4
     movdqa    [edx], xmm0
@@ -5806,7 +5807,6 @@ void ComputeCumulativeSumRow_SSE2(const uint8* row, int32* cumsum,
     mov        edx, cumsum
     mov        esi, previous_cumsum
     mov        ecx, width
-    sub        esi, edx
     pxor       xmm0, xmm0
     pxor       xmm1, xmm1
 
@@ -5833,19 +5833,20 @@ void ComputeCumulativeSumRow_SSE2(const uint8* row, int32* cumsum,
     punpckhwd  xmm5, xmm1
 
     paddd      xmm0, xmm2
-    movdqa     xmm2, [edx + esi]  // previous row above.
+    movdqa     xmm2, [esi]  // previous row above.
     paddd      xmm2, xmm0
 
     paddd      xmm0, xmm3
-    movdqa     xmm3, [edx + esi + 16]
+    movdqa     xmm3, [esi + 16]
     paddd      xmm3, xmm0
 
     paddd      xmm0, xmm4
-    movdqa     xmm4, [edx + esi + 32]
+    movdqa     xmm4, [esi + 32]
     paddd      xmm4, xmm0
 
     paddd      xmm0, xmm5
-    movdqa     xmm5, [edx + esi + 48]
+    movdqa     xmm5, [esi + 48]
+    lea        esi, [esi + 64]
     paddd      xmm5, xmm0
 
     movdqa     [edx], xmm2
@@ -5869,7 +5870,8 @@ void ComputeCumulativeSumRow_SSE2(const uint8* row, int32* cumsum,
     punpcklbw  xmm2, xmm1
     punpcklwd  xmm2, xmm1
     paddd      xmm0, xmm2
-    movdqu     xmm2, [edx + esi]
+    movdqu     xmm2, [esi]
+    lea        esi, [esi + 16]
     paddd      xmm2, xmm0
     movdqu     [edx], xmm2
     lea        edx, [edx + 16]