Implement SSE2-based implementations of the morphology filters (dilate &

src/effects/SkMorphologyImageFilter.cpp

 /*
  * Copyright 2012 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 "SkMorphologyImageFilter.h"
 #include "SkBitmap.h"
 #include "SkColorPriv.h"
 #include "SkFlattenableBuffers.h"
 #include "SkRect.h"
+#include "SkMorphology_opts.h"
 #if SK_SUPPORT_GPU
 #include "GrContext.h"
 #include "GrTexture.h"
 #include "GrTBackendEffectFactory.h"
 #include "gl/GrGLEffect.h"
 #include "effects/Gr1DKernelEffect.h"
 #include "SkImageFilterUtils.h"
 #endif
 
 SkMorphologyImageFilter::SkMorphologyImageFilter(SkFlattenableReadBuffer& buffer)
@@ skipping 44 matching lines @@
             up += srcStrideY;
         }
         if (x >= radius) src += srcStrideX;
         if (x + radius < width - 1) upperSrc += srcStrideX;
         dst += dstStrideX;
     }
 }
 
 static void erodeX(const SkBitmap& src, SkBitmap* dst, int radiusX, const SkIRect& bounds)
 {
-    erode(src.getAddr32(bounds.left(), bounds.top()), dst->getAddr32(0, 0),
-          radiusX, bounds.width(), bounds.height(),
-          1, src.rowBytesAsPixels(), 1, dst->rowBytesAsPixels());
+    if (SkMorphologyProc erodeXProc = SkErodeXGetPlatformProc()) {

[reed1, 2013/10/30 15:15:04]

can we change either the platformproc or your erode function to have the same
signature? That would make the call sites easier to read/confirm that we're
passing the right parameters...

e.g.

xproc = SkErodeXGetPlatformProc();
if (NULL == xproc) {
    xproc = erode_x;
}
xproc(...);

[Stephen White, 2013/10/30 19:47:17]

Done.

+        (*erodeXProc)(src.getAddr32(bounds.left(), bounds.top()),
+                      dst->getAddr32(0, 0), radiusX,
+                      bounds.width(), bounds.height(),
+                      src.rowBytesAsPixels(),
+                      dst->rowBytesAsPixels());
+    } else {
+        erode(src.getAddr32(bounds.left(), bounds.top()), dst->getAddr32(0, 0),
+              radiusX, bounds.width(), bounds.height(),
+              1, src.rowBytesAsPixels(), 1, dst->rowBytesAsPixels());
+    }
 }
 
 static void erodeY(const SkBitmap& src, SkBitmap* dst, int radiusY, const SkIRect& bounds)
 {
-    erode(src.getAddr32(bounds.left(), bounds.top()), dst->getAddr32(0, 0),
-          radiusY, bounds.height(), bounds.width(),
-          src.rowBytesAsPixels(), 1, dst->rowBytesAsPixels(), 1);
+    if (SkMorphologyProc erodeYProc = SkErodeYGetPlatformProc()) {
+        (*erodeYProc)(src.getAddr32(bounds.left(), bounds.top()),
+                      dst->getAddr32(0, 0),
+                      radiusY,
+                      bounds.width(), bounds.height(),
+                      src.rowBytesAsPixels(),
+                      dst->rowBytesAsPixels());
+    } else {
+        erode(src.getAddr32(bounds.left(), bounds.top()), dst->getAddr32(0, 0),
+              radiusY, bounds.height(), bounds.width(),
+              src.rowBytesAsPixels(), 1, dst->rowBytesAsPixels(), 1);
+    }
 }
 
 static void dilate(const SkPMColor* src, SkPMColor* dst,
                    int radius, int width, int height,
                    int srcStrideX, int srcStrideY,
                    int dstStrideX, int dstStrideY)
 {
     radius = SkMin32(radius, width - 1);
     const SkPMColor* upperSrc = src + radius * srcStrideX;
     for (int x = 0; x < width; ++x) {
@@ skipping 18 matching lines @@
             up += srcStrideY;
         }
         if (x >= radius) src += srcStrideX;
         if (x + radius < width - 1) upperSrc += srcStrideX;
         dst += dstStrideX;
     }
 }
 
 static void dilateX(const SkBitmap& src, SkBitmap* dst, int radiusX, const SkIRect& bounds)
 {
-    dilate(src.getAddr32(bounds.left(), bounds.top()), dst->getAddr32(0, 0),
-           radiusX, bounds.width(), bounds.height(),
-           1, src.rowBytesAsPixels(), 1, dst->rowBytesAsPixels());
+    if (SkMorphologyProc erodeYProc = SkDilateXGetPlatformProc()) {
+        (*erodeYProc)(src.getAddr32(bounds.left(), bounds.top()),
+                      dst->getAddr32(0, 0), radiusX,
+                      bounds.width(), bounds.height(),
+                      src.rowBytesAsPixels(),
+                      dst->rowBytesAsPixels());
+    } else {
+        dilate(src.getAddr32(bounds.left(), bounds.top()), dst->getAddr32(0, 0),
+               radiusX, bounds.width(), bounds.height(),
+               1, src.rowBytesAsPixels(), 1, dst->rowBytesAsPixels());
+    }
 }
 
 static void dilateY(const SkBitmap& src, SkBitmap* dst, int radiusY, const SkIRect& bounds)
 {
-    dilate(src.getAddr32(bounds.left(), bounds.top()), dst->getAddr32(0, 0),
-           radiusY, bounds.height(), bounds.width(),
-           src.rowBytesAsPixels(), 1, dst->rowBytesAsPixels(), 1);
+    if (SkMorphologyProc dilateYProc = SkDilateYGetPlatformProc()) {
+        (*dilateYProc)(src.getAddr32(bounds.left(), bounds.top()),
+                       dst->getAddr32(0, 0),
+                       radiusY,
+                       bounds.width(), bounds.height(),
+                       src.rowBytesAsPixels(),
+                       dst->rowBytesAsPixels());
+    } else {
+        dilate(src.getAddr32(bounds.left(), bounds.top()), dst->getAddr32(0, 0),
+               radiusY, bounds.height(), bounds.width(),
+               src.rowBytesAsPixels(), 1, dst->rowBytesAsPixels(), 1);
+    }
 }
 
 bool SkErodeImageFilter::onFilterImage(Proxy* proxy,
                                        const SkBitmap& source, const SkMatrix& ctm,
                                        SkBitmap* dst, SkIPoint* offset) {
     SkBitmap src = source;
     if (getInput(0) && !getInput(0)->filterImage(proxy, source, ctm, &src, offset)) {
         return false;
     }
 
@@ skipping 440 matching lines @@
 
     if (!apply_morphology(input, bounds, GrMorphologyEffect::kErode_MorphologyType, radius(), result)) {
         return false;
     }
     offset->fX += bounds.left();
     offset->fY += bounds.top();
     return true;
 }
 
 #endif