Add SK_PREFETCH and use in SkBlurImageFilter.

src/effects/SkBlurImageFilter.cpp

 /*
  * Copyright 2011 The Android Open Source Project
  *
  * Use of this source code is governed by a BSD-style license that can be
  * found in the LICENSE file.
  */
 
 #include "SkBitmap.h"
 #include "SkBlurImageFilter.h"
 #include "SkColorPriv.h"
@@ skipping 126 matching lines @@
                 sumB -= SkGetPackedB32(l);
             }
             if (y + bottomOffset + 1 < height) {
                 SkColor r = *(sptr + (bottomOffset + 1) * srcStride);
                 sumA += SkGetPackedA32(r);
                 sumR += SkGetPackedR32(r);
                 sumG += SkGetPackedG32(r);
                 sumB += SkGetPackedB32(r);
             }
             sptr += srcStride;
+            // The next leading pixel seems to be too hard to predict.  Hint the fetch.
+            SK_PREFETCH(sptr + (bottomOffset + 1) * srcStride);
             dptr += dstStride;
         }
     }
 }
 
 static void getBox3Params(SkScalar s, int *kernelSize, int* kernelSize3, int *lowOffset,
                           int *highOffset)
 {
     float pi = SkScalarToFloat(SK_ScalarPI);
     int d = static_cast<int>(floorf(SkScalarToFloat(s) * 3.0f * sqrtf(2.0f * pi) / 4.0f + 0.5f));
@@ skipping 96 matching lines @@
                                                              fSigma.width(),
                                                              fSigma.height()));
     offset->fX += rect.fLeft;
     offset->fY += rect.fTop;
     return SkImageFilterUtils::WrapTexture(tex, rect.width(), rect.height(), result);
 #else
     SkDEBUGFAIL("Should not call in GPU-less build");
     return false;
 #endif
 }