Plus xfermode using Sk4px.

src/core/Sk4px.h

Index: src/core/Sk4px.h
diff --git a/src/core/Sk4px.h b/src/core/Sk4px.h
index 3d2a8e336222c893f345d68af2d33557e2dc7640..830e60a156567c2859a382a799926077cd6522ef 100644
--- a/src/core/Sk4px.h
+++ b/src/core/Sk4px.h
@@ -34,6 +34,12 @@ public:
 
         // Pack the top byte of each component back down into 4 SkPMColors.
         Sk4px addNarrowHi(const Sk16h&) const;
+
+        Sk4px div255TruncNarrow() const { return this->addNarrowHi(*this >> 8); }
+        Sk4px div255RoundNarrow() const {
+            return Sk4px::Wide(*this + Sk16h(128)).div255TruncNarrow();
+        }
+
     private:
         typedef Sk16h INHERITED;
     };
@@ -73,6 +79,72 @@ public:
         }
     }
 
+    // As above, but with dst4' = fn(dst4, src4).
+    template <typename Fn>
+    static void MapDstSrc(int count, SkPMColor* dst, const SkPMColor* src, Fn fn) {
+        while (count > 0) {
+            if (count >= 8) {
+                Sk4px dst0 = fn(Load4(dst+0), Load4(src+0)),
+                      dst4 = fn(Load4(dst+4), Load4(src+4));
+                dst0.store4(dst+0);
+                dst4.store4(dst+4);
+                dst += 8; src += 8; count -= 8;
+                continue;  // Keep our stride at 8 pixels as long as possible.
+            }
+            SkASSERT(count <= 7);
+            if (count >= 4) {
+                fn(Load4(dst), Load4(src)).store4(dst);
+                dst += 4; src += 4; count -= 4;
+            }
+            if (count >= 2) {
+                fn(Load2(dst), Load2(src)).store2(dst);
+                dst += 2; src += 2; count -= 2;
+            }
+            if (count >= 1) {
+                fn(Load1(dst), Load1(src)).store1(dst);
+            }
+            break;
+        }
+    }
+
+    // As above, but with dst4' = fn(dst4, src4, alpha4).
+    template <typename Fn>
+    static void MapDstSrcAlpha(
+            int count, SkPMColor* dst, const SkPMColor* src, const SkAlpha* a, Fn fn) {
+        // TODO: find a terser / faster way to construct Sk16b alphas.
+        while (count > 0) {
+            if (count >= 8) {
+                Sk16b alpha0(a[0],a[0],a[0],a[0], a[1],a[1],a[1],a[1],
+                             a[2],a[2],a[2],a[2], a[3],a[3],a[3],a[3]),
+                      alpha4(a[4],a[4],a[4],a[4], a[5],a[5],a[5],a[5],
+                             a[6],a[6],a[6],a[6], a[7],a[7],a[7],a[7]);
+                Sk4px dst0 = fn(Load4(dst+0), Load4(src+0), alpha0),
+                      dst4 = fn(Load4(dst+4), Load4(src+4), alpha4);
+                dst0.store4(dst+0);
+                dst4.store4(dst+4);
+                dst += 8; src += 8; a += 8; count -= 8;
+                continue;  // Keep our stride at 8 pixels as long as possible.
+            }
+            SkASSERT(count <= 7);
+            if (count >= 4) {
+                Sk16b alpha(a[0],a[0],a[0],a[0], a[1],a[1],a[1],a[1],
+                            a[2],a[2],a[2],a[2], a[3],a[3],a[3],a[3]);
+                fn(Load4(dst), Load4(src), alpha).store4(dst);
+                dst += 4; src += 4; a += 4; count -= 4;
+            }
+            if (count >= 2) {
+                Sk16b alpha(a[0],a[0],a[0],a[0], a[1],a[1],a[1],a[1], 0,0,0,0, 0,0,0,0);
+                fn(Load2(dst), Load2(src), alpha).store2(dst);
+                dst += 2; src += 2; a += 2; count -= 2;
+            }
+            if (count >= 1) {
+                Sk16b alpha(a[0],a[0],a[0],a[0], 0,0,0,0, 0,0,0,0, 0,0,0,0);
+                fn(Load1(dst), Load1(src), alpha).store1(dst);
+            }
+            break;
+        }
+    }
+
 private:
     typedef Sk16b INHERITED;
 };