Add 8888 fast SrcOver mode.

src/core/SkLinearBitmapPipeline_core.h

 /*
  * Copyright 2016 Google Inc.
  *
  * Use of this source code is governed by a BSD-style license that can be
  * found in the LICENSE file.
  */
 
 #ifndef SkLinearBitmapPipeline_core_DEFINED
 #define SkLinearBitmapPipeline_core_DEFINED
 
 #include <cmath>
+#include "SkNx.h"
 
 // New bilerp strategy:
 // Pass through on bilerpList4 and bilerpListFew (analogs to pointList), introduce bilerpEdge
 // which takes 4 points. If the sample spans an edge, then break it into a bilerpEdge. Bilerp
 // span then becomes a normal span except in special cases where an extra Y is given. The bilerp
 // need to stay single point calculations until the tile layer.
 // TODO:
 //  - edge span predicate.
 //  - introduce new point API
 //  - Add tile for new api.
@@ skipping 159 matching lines @@
         xs = xs + fourDx;
         count -= 4;
     }
     if (count > 0) {
         stage->pointListFew(count, xs, ys);
     }
 }
 }  // namespace
 
 #endif // SkLinearBitmapPipeline_core_DEFINED