ARGB Multiply

BUG=none
TEST=none
Review URL: https://webrtc-codereview.appspot.com/1028010

git-svn-id: http://libyuv.googlecode.com/svn/trunk@537 16f28f9a-4ce2-e073-06de-1de4eb20be90

README.chromium

 Name: libyuv
 URL: http://code.google.com/p/libyuv/
-Version: 535
+Version: 537
 License: BSD
 License File: LICENSE
 

include/libyuv/row.h

@@ -118,6 +118,7 @@ extern "C" {
 // TODO(fbarchard): Port to gcc.
 #if !defined(YUV_DISABLE_ASM) && defined(_M_IX86)
 #define HAS_ARGBCOLORTABLEROW_X86
+#define HAS_ARGBMULTIPLYROW_SSE2
 #endif
 
 // The following are Yasm x86 only.
@@ -1278,6 +1279,7 @@ void ARGBInterpolateRow_SSSE3(uint8* dst_argb, const uint8* src_argb,
 void ARGBInterpolateRow_NEON(uint8* dst_argb, const uint8* src_argb,
                              ptrdiff_t src_stride_argb, int dst_width,
                              int source_y_fraction);
+void ARGBMultiplyRow_SSE2(const uint8* src_argb, uint8* dst_argb, int width);
 
 #ifdef __cplusplus
 }  // extern "C"

include/libyuv/version.h

@@ -11,6 +11,6 @@
 #ifndef INCLUDE_LIBYUV_VERSION_H_  // NOLINT
 #define INCLUDE_LIBYUV_VERSION_H_
 
-#define LIBYUV_VERSION 536
+#define LIBYUV_VERSION 537
 
 #endif  // INCLUDE_LIBYUV_VERSION_H_  NOLINT

source/row_win.cc

@@ -4160,6 +4160,41 @@ void ARGBShadeRow_SSE2(const uint8* src_argb, uint8* dst_argb, int width,
 }
 #endif  // HAS_ARGBSHADEROW_SSE2
 
+#ifdef HAS_ARGBMULTIPLYROW_SSE2
+// Multiple 2 rows of ARGB pixels together, 4 pixels at a time.
+// Aligned to 16 bytes.
+__declspec(naked) __declspec(align(16))
+void ARGBMultiplyRow_SSE2(const uint8* src_argb, uint8* dst_argb, int width) {
+  __asm {
+    mov        eax, [esp + 4]   // src_argb
+    mov        edx, [esp + 8]   // dst_argb
+    mov        ecx, [esp + 12]  // width
+    pxor       xmm5, xmm5  // constant 0
+    sub        edx, eax
+
+    align      16
+ convertloop:
+    movdqa     xmm0, [eax]      // read 4 pixels
+    movdqa     xmm2, [eax + edx]  // read 4 dest pixels
+    movdqa     xmm1, xmm0
+    movdqa     xmm3, xmm2
+    punpcklbw  xmm0, xmm0       // first 2
+    punpckhbw  xmm1, xmm1       // next 2
+    punpcklbw  xmm2, xmm5       // first 2
+    punpckhbw  xmm3, xmm5       // next 2
+    pmulhuw    xmm0, xmm2       // argb * value
+    pmulhuw    xmm1, xmm3       // argb * value
+    packuswb   xmm0, xmm1
+    sub        ecx, 4
+    movdqa     [eax + edx], xmm0
+    lea        eax, [eax + 16]
+    jg         convertloop
+
+    ret
+  }
+}
+#endif  // HAS_ARGBMULTIPLYROW_SSE2
+
 #ifdef HAS_CUMULATIVESUMTOAVERAGEROW_SSE2
 // Consider float CumulativeSum.
 // Consider calling CumulativeSum one row at time as needed.