Working plumb of image scaling:

src/opts/SkBitmapFilter_opts_SSE2.cpp

 /*
  * Copyright 2013 Google Inc.
  *
  * Use of this source code is governed by a BSD-style license that can be
  * found in the LICENSE file.
  */
 
 #include "SkBitmapProcState.h"
 #include "SkBitmap.h"
 #include "SkColor.h"
@@ skipping 37 matching lines @@
 int debug_y = 255;
 
 void highQualityFilter_SSE2(const SkBitmapProcState& s, int x, int y,
                         SkPMColor* SK_RESTRICT colors, int count) {
 
     const int maxX = s.fBitmap->width() - 1;
     const int maxY = s.fBitmap->height() - 1;
 
     while (count-- > 0) {
         SkPoint srcPt;
-        s.fInvProc(*s.fInvMatrix, SkIntToScalar(x),
+        s.fInvProc(s.fInvMatrix, SkIntToScalar(x),
                     SkIntToScalar(y), &srcPt);
         srcPt.fX -= SK_ScalarHalf;
         srcPt.fY -= SK_ScalarHalf;
 
         int sx = SkScalarFloorToInt(srcPt.fX);
         int sy = SkScalarFloorToInt(srcPt.fY);
 
         __m128 weight = _mm_setzero_ps();
         __m128 accum = _mm_setzero_ps();
 
         int y0 = SkTMax(0, int(ceil(sy-s.getBitmapFilter()->width() + 0.5f)));
         int y1 = SkTMin(maxY, int(floor(sy+s.getBitmapFilter()->width() + 0.5f)));
         int x0 = SkTMax(0, int(ceil(sx-s.getBitmapFilter()->width() + 0.5f)));
         int x1 = SkTMin(maxX, int(floor(sx+s.getBitmapFilter()->width() + 0.5f)));
 
         for (int src_y = y0; src_y <= y1; src_y++) {
-            float yweight = s.getBitmapFilter()->lookupFloat( (srcPt.fY - src_y) );
+            float yweight = SkScalarToFloat(s.getBitmapFilter()->lookupScalar(srcPt.fY - src_y));
 
             for (int src_x = x0; src_x <= x1 ; src_x++) {
-                float xweight = s.getBitmapFilter()->lookupFloat( (srcPt.fX - src_x) );
+                float xweight = SkScalarToFloat(s.getBitmapFilter()->lookupScalar(srcPt.fX - src_x));
 
                 float combined_weight = xweight * yweight;
 
                 SkPMColor color = *s.fBitmap->getAddr32(src_x, src_y);
 
                 __m128i c = _mm_cvtsi32_si128( color );
                 c = _mm_unpacklo_epi8(c, _mm_setzero_si128());
                 c = _mm_unpacklo_epi16(c, _mm_setzero_si128());
 
                 __m128 cfloat = _mm_cvtepi32_ps( c );
@@ skipping 22 matching lines @@
         x++;
     }
 }
 
 void highQualityFilter_ScaleOnly_SSE2(const SkBitmapProcState &s, int x, int y,
                              SkPMColor *SK_RESTRICT colors, int count) {
     const int maxX = s.fBitmap->width() - 1;
     const int maxY = s.fBitmap->height() - 1;
 
     SkPoint srcPt;
-    s.fInvProc(*s.fInvMatrix, SkIntToScalar(x),
+    s.fInvProc(s.fInvMatrix, SkIntToScalar(x),
                 SkIntToScalar(y), &srcPt);
     srcPt.fY -= SK_ScalarHalf;
     int sy = SkScalarFloorToInt(srcPt.fY);
 
     int y0 = SkTMax(0, int(ceil(sy-s.getBitmapFilter()->width() + 0.5f)));
     int y1 = SkTMin(maxY, int(floor(sy+s.getBitmapFilter()->width() + 0.5f)));
 
     while (count-- > 0) {
         srcPt.fX -= SK_ScalarHalf;
         srcPt.fY -= SK_ScalarHalf;
 
         int sx = SkScalarFloorToInt(srcPt.fX);
 
         float weight = 0;
         __m128 accum = _mm_setzero_ps();
 
         int x0 = SkTMax(0, int(ceil(sx-s.getBitmapFilter()->width() + 0.5f)));
         int x1 = SkTMin(maxX, int(floor(sx+s.getBitmapFilter()->width() + 0.5f)));
 
         for (int src_y = y0; src_y <= y1; src_y++) {
-            float yweight = s.getBitmapFilter()->lookupFloat( (srcPt.fY - src_y) );
+            float yweight = SkScalarToFloat(s.getBitmapFilter()->lookupScalar(srcPt.fY - src_y));
 
             for (int src_x = x0; src_x <= x1 ; src_x++) {
-                float xweight = s.getBitmapFilter()->lookupFloat( (srcPt.fX - src_x) );
+                float xweight = SkScalarToFloat(s.getBitmapFilter()->lookupScalar(srcPt.fX - src_x));
 
                 float combined_weight = xweight * yweight;
 
                 SkPMColor color = *s.fBitmap->getAddr32(src_x, src_y);
 
                 __m128 c = _mm_set_ps((float)SkGetPackedB32(color),
                                       (float)SkGetPackedG32(color),
                                       (float)SkGetPackedR32(color),
                                       (float)SkGetPackedA32(color));
 
@@ skipping 12 matching lines @@
         _mm_storeu_ps(localResult, accum);
         int a = SkClampMax(int(localResult[0]), 255);
         int r = SkClampMax(int(localResult[1]), a);
         int g = SkClampMax(int(localResult[2]), a);
         int b = SkClampMax(int(localResult[3]), a);
 
         *colors++ = SkPackARGB32(a, r, g, b);
 
         x++;
 
-        s.fInvProc(*s.fInvMatrix, SkIntToScalar(x),
+        s.fInvProc(s.fInvMatrix, SkIntToScalar(x),
                     SkIntToScalar(y), &srcPt);
 
     }
 }