Unroll C version of YUV blender for improved performance.

source/row_common.cc

 /*
  *  Copyright 2011 The LibYuv Project Authors. All rights reserved.
  *
  *  Use of this source code is governed by a BSD-style license
  *  that can be found in the LICENSE file in the root of the source
  *  tree. An additional intellectual property rights grant can be found
  *  in the file PATENTS. All contributing project authors may
  *  be found in the AUTHORS file in the root of the source tree.
  */
 
@@ skipping 1999 matching lines @@
     uint32 bg = src_argb1[1];
     uint32 br = src_argb1[2];
     dst_argb[0] = BLEND(fb, bb, a);
     dst_argb[1] = BLEND(fg, bg, a);
     dst_argb[2] = BLEND(fr, br, a);
     dst_argb[3] = 255u;
   }
 }
 #undef BLEND
 
+#define UBLEND(f, b, a) (((a) * f) + ((255 - a) * b) + 255) >> 8
 void BlendPlaneRow_C(const uint8* src0, const uint8* src1,
                      const uint8* alpha, uint8* dst, int width) {
   int x;
-  for (x = 0; x < width; ++x) {
-    uint32 f = *src0++;
-    uint32 b = *src1++;
-    uint32 a = *alpha++;
-    *dst++ = (((a) * f) + ((255 - a) * b) + 255) >> 8;
+  for (x = 0; x < width - 1; x += 2) {
+    dst[0] = UBLEND(src0[0], src1[0], alpha[0]);
+    dst[1] = UBLEND(src0[1], src1[1], alpha[1]);
+    src0 += 2;
+    src1 += 2;
+    alpha += 2;
+    dst += 2;
+  }
+  if (width & 1) {
+    dst[0] = UBLEND(src0[0], src1[0], alpha[0]);
   }
 }
+#undef UBLEND
 
 #define ATTENUATE(f, a) (a | (a << 8)) * (f | (f << 8)) >> 24
 
 // Multiply source RGB by alpha and store to destination.
 // This code mimics the SSSE3 version for better testability.
 void ARGBAttenuateRow_C(const uint8* src_argb, uint8* dst_argb, int width) {
   int i;
   for (i = 0; i < width - 1; i += 2) {
     uint32 b = src_argb[0];
     uint32 g = src_argb[1];
@@ skipping 583 matching lines @@
     dst_rgb565 += twidth * 2;
     width -= twidth;
   }
 }
 #endif
 
 #ifdef __cplusplus
 }  // extern "C"
 }  // namespace libyuv
 #endif