Exclusion and Difference modes using Sk4f

src/core/SkXfermode.cpp

Index: src/core/SkXfermode.cpp
diff --git a/src/core/SkXfermode.cpp b/src/core/SkXfermode.cpp
index 92743c2fddeddd86d62dc9df14f358c5ee22cdc4..d78dcd8d4932041a28ec9c19575feff4c83daadf 100644
--- a/src/core/SkXfermode.cpp
+++ b/src/core/SkXfermode.cpp
@@ -1201,6 +1201,13 @@ static Sk4f clamp_0_255(const Sk4f& value) {
     return Sk4f::Max(Sk4f(0), Sk4f::Min(Sk4f(255), value));
 }
 
+// return a swizzle of a | rgb
+static Sk4f set_a_rgb(const Sk4f& a, const Sk4f& rgb) {
+    SkPMFloat pma = a;
+    SkPMFloat pmc = rgb;
+    return SkPMFloat(pma.a(), pmc.r(), pmc.g(), pmc.b());

[mtklein, 2015/04/03 13:23:58]

Boy, this is annoying.  We can make this operation faster (1 instruction) if we
stored alpha in the low position, but then we'd not get the SkPMColor /
SkPMFloat color-order alignment that makes conversion fast.

[mtklein, 2015/04/03 13:35:47]

Well, maybe it's not so annoying.  On x86 this can be done in 2 shuffle
instructions, and it looks like Clang's doing that just fine.  Probably not
worth fussing over.

+}
+
 /**
  *  Some modes can, due to very slight numerical error, generate "invalid" pmcolors...
  *
@@ -1311,6 +1318,34 @@ struct Multiply4f {
     static const SkXfermode::Mode kMode = SkXfermode::kMultiply_Mode;
 };
 
+struct Difference4f {
+    static SkPMFloat Xfer(const SkPMFloat& src, const SkPMFloat& dst) {
+        const Sk4f inv255(gInv255);
+        Sk4f sa = Sk4f(src.a());
+        Sk4f da = Sk4f(dst.a());
+        Sk4f sc = src;
+        Sk4f dc = dst;
+        Sk4f min = Sk4f::Min(sc * da, dc * sa) * inv255;
+        Sk4f ra = sc + dc - min;
+        return check_as_pmfloat(set_a_rgb(ra, ra - min));
+    }
+    static const bool kFoldCoverageIntoSrcAlpha = false;
+    static const SkXfermode::Mode kMode = SkXfermode::kDifference_Mode;
+};
+
+struct Exclusion4f {
+    static SkPMFloat Xfer(const SkPMFloat& src, const SkPMFloat& dst) {
+        const Sk4f inv255(gInv255);
+        Sk4f sc = src;
+        Sk4f dc = dst;
+        Sk4f prod = sc * dc * inv255;
+        Sk4f ra = sc + dc - prod;
+        return check_as_pmfloat(set_a_rgb(ra, ra - prod));
+    }
+    static const bool kFoldCoverageIntoSrcAlpha = false;
+    static const SkXfermode::Mode kMode = SkXfermode::kExclusion_Mode;
+};
+
 template <typename ProcType>
 class SkT4fXfermode : public SkProcCoeffXfermode {
 public:
@@ -1445,6 +1480,12 @@ SkXfermode* create_mode(int iMode) {
         case SkXfermode::kMultiply_Mode:
             xfer = SkT4fXfermode<Multiply4f>::Create(rec);
             break;
+        case SkXfermode::kDifference_Mode:
+            xfer = SkT4fXfermode<Difference4f>::Create(rec);
+            break;
+        case SkXfermode::kExclusion_Mode:
+            xfer = SkT4fXfermode<Exclusion4f>::Create(rec);
+            break;
         default:
             break;
     }