create shaderproc for nofilter-opaque-dx

src/core/SkBitmapProcState.cpp

Index: src/core/SkBitmapProcState.cpp
diff --git a/src/core/SkBitmapProcState.cpp b/src/core/SkBitmapProcState.cpp
index 2a449d6d8fa8e7732732bd9046517eb2a2ddb57d..4528d64f8f910b6338431f2d18d1f8e424c43856 100644
--- a/src/core/SkBitmapProcState.cpp
+++ b/src/core/SkBitmapProcState.cpp
@@ -29,6 +29,7 @@ extern void  SI8_opaque_D32_filter_DX_neon(const SkBitmapProcState&, const uint3
 extern void  SI8_opaque_D32_filter_DX_shaderproc_neon(const SkBitmapProcState&, int, int, uint32_t*, int);
 extern void  Clamp_SI8_opaque_D32_filter_DX_shaderproc_neon(const SkBitmapProcState&, int, int, uint32_t*, int);
 #endif
+extern void  Clamp_S32_opaque_D32_nofilter_DX_shaderproc(const SkBitmapProcState&, int, int, uint32_t*, int);
 
 #define   NAME_WRAP(x)  x
 #include "SkBitmapProcState_filter.h"
@@ -589,6 +590,8 @@ bool SkBitmapProcState::chooseProcs(const SkMatrix& inv, const SkPaint& paint) {
             }
         } else if (SK_ARM_NEON_WRAP(SI8_opaque_D32_filter_DX) == fSampleProc32 && clampClamp) {
             fShaderProc32 = SK_ARM_NEON_WRAP(Clamp_SI8_opaque_D32_filter_DX_shaderproc);
+        } else if (SK_ARM_NEON_WRAP(S32_opaque_D32_nofilter_DX) == fSampleProc32 && clampClamp) {

[djsollen, 2014/10/23 19:00:33]

compile fails because S32_opaque_D32_nofilter_DX_neon doesn't exist

[reed1, 2014/10/23 19:13:56]

Done.

+            fShaderProc32 = Clamp_S32_opaque_D32_nofilter_DX_shaderproc;
         }
 
         if (NULL == fShaderProc32) {
@@ -1019,3 +1022,54 @@ int SkBitmapProcState::maxCountForBufferSize(size_t bufferSize) const {
 
     return size;
 }
+
+///////////////////////
+
+void  Clamp_S32_opaque_D32_nofilter_DX_shaderproc(const SkBitmapProcState& s, int x, int y,
+                                                  SkPMColor* SK_RESTRICT dst, int count) {
+    SkASSERT((s.fInvType & ~(SkMatrix::kTranslate_Mask |
+                             SkMatrix::kScale_Mask)) == 0);
+
+    const unsigned maxX = s.fBitmap->width() - 1;
+    SkFractionalInt fx;
+    int dstY;
+    {
+        SkPoint pt;
+        s.fInvProc(s.fInvMatrix, SkIntToScalar(x) + SK_ScalarHalf, SkIntToScalar(y) + SK_ScalarHalf,
+                   &pt);
+        fx = SkScalarToFractionalInt(pt.fY);
+        const unsigned maxY = s.fBitmap->height() - 1;
+        dstY = SkClampMax(SkFractionalIntToInt(fx), maxY);
+        fx = SkScalarToFractionalInt(pt.fX);
+    }
+
+    const SkPMColor* SK_RESTRICT src = s.fBitmap->getAddr32(0, dstY);
+    const SkFractionalInt dx = s.fInvSxFractionalInt;
+
+    if ((unsigned)(fx >> 48) <= maxX && (unsigned)((fx + dx * (count - 1)) >> 48) <= maxX) {

[mtklein, 2014/10/23 19:05:45]

Can't hurt to make this a little more readable?

if (maxX >= (unsigned)SkFractionalIntToInt(fx) && 
    maxX >= (unsigned)SkFractionalIntToInt(fx + dx * (count-1))) {
  // No need to clamp anywhere in our x range.
  ...
}

[reed1, 2014/10/23 19:13:56]

Done.

+        int count4 = count >> 2;
+        for (int i = 0; i < count4; ++i) {
+            SkPMColor src0 = src[SkFractionalIntToInt(fx)]; fx += dx;
+            SkPMColor src1 = src[SkFractionalIntToInt(fx)]; fx += dx;
+            SkPMColor src2 = src[SkFractionalIntToInt(fx)]; fx += dx;
+            SkPMColor src3 = src[SkFractionalIntToInt(fx)]; fx += dx;
+            dst[0] = src0;
+            dst[1] = src1;
+            dst[2] = src2;
+            dst[3] = src3;
+            dst += 4;
+        }
+        for (int i = (count4 << 2); i < count; ++i) {
+            unsigned index = SkFractionalIntToInt(fx);
+            SkASSERT(index <= maxX);
+            *dst++ = src[index];
+            fx += dx;
+        }
+    } else {
+        for (int i = 0; i < count; ++i) {
+            dst[i] = src[SkClampMax(SkFractionalIntToInt(fx), maxX)];
+            fx += dx;
+        }
+    }
+}
+