Revert of Normalize SkXfermode_opts.h argument order as d,s[,aa].

src/opts/SkXfermode_opts.h

 /*
  * Copyright 2015 Google Inc.
  *
  * Use of this source code is governed by a BSD-style license that can be
  * found in the LICENSE file.
  */
 
 #ifndef Sk4pxXfermode_DEFINED
 #define Sk4pxXfermode_DEFINED
 
 #include "Sk4px.h"
 #include "SkPMFloat.h"
 #include "SkXfermode_proccoeff.h"
 
 namespace SK_OPTS_NS {
 
 // Most xfermodes can be done most efficiently 4 pixels at a time in 8 or 16-bit fixed point.
-#define XFERMODE(Name) static Sk4px SK_VECTORCALL Name(Sk4px d, Sk4px s)
+#define XFERMODE(Name) static Sk4px SK_VECTORCALL Name(Sk4px s, Sk4px d)
 
 XFERMODE(Clear) { return Sk4px::DupPMColor(0); }
 XFERMODE(Src)   { return s; }
 XFERMODE(Dst)   { return d; }
 XFERMODE(SrcIn)   { return     s.approxMulDiv255(d.alphas()      ); }
 XFERMODE(SrcOut)  { return     s.approxMulDiv255(d.alphas().inv()); }
 XFERMODE(SrcOver) { return s + d.approxMulDiv255(s.alphas().inv()); }
-XFERMODE(DstIn)   { return SrcIn  (s,d); }
-XFERMODE(DstOut)  { return SrcOut (s,d); }
-XFERMODE(DstOver) { return SrcOver(s,d); }
+XFERMODE(DstIn)   { return SrcIn  (d,s); }
+XFERMODE(DstOut)  { return SrcOut (d,s); }
+XFERMODE(DstOver) { return SrcOver(d,s); }
 
 // [ S * Da + (1 - Sa) * D]
 XFERMODE(SrcATop) { return (s * d.alphas() + d * s.alphas().inv()).div255(); }
-XFERMODE(DstATop) { return SrcATop(s,d); }
+XFERMODE(DstATop) { return SrcATop(d,s); }
 //[ S * (1 - Da) + (1 - Sa) * D ]
 XFERMODE(Xor) { return (s * d.alphas().inv() + d * s.alphas().inv()).div255(); }
 // [S + D ]
 XFERMODE(Plus) { return s.saturatedAdd(d); }
 // [S * D ]
 XFERMODE(Modulate) { return s.approxMulDiv255(d); }
 // [S + D - S * D]
 XFERMODE(Screen) {
     // Doing the math as S + (1-S)*D or S + (D - S*D) means the add and subtract can be done
     // in 8-bit space without overflow.  S + (1-S)*D is a touch faster because inv() is cheap.
@@ skipping 29 matching lines @@
     auto isLite = ((sa-s) < s).widenLoHi();
 
     auto lite = sa*da - ((da-d)*(sa-s) << 1),
          dark = s*d << 1,
          both = s*da.inv() + d*sa.inv();
 
     auto alphas = srcover;
     auto colors = (both + isLite.thenElse(lite, dark)).div255();
     return alphas.zeroColors() + colors.zeroAlphas();
 }
-XFERMODE(Overlay) { return HardLight(s,d); }
+XFERMODE(Overlay) { return HardLight(d,s); }
 
 XFERMODE(Darken) {
     auto sa = s.alphas(),
          da = d.alphas();
 
     auto sda = (s*da).div255(),
          dsa = (d*sa).div255();
 
     auto srcover = s + (d * sa.inv()).div255(),
          dstover = d + (s * da.inv()).div255();
@@ skipping 10 matching lines @@
 
     auto srcover = s + (d * sa.inv()).div255(),
          dstover = d + (s * da.inv()).div255();
     auto alphas = srcover,
          colors = (dsa < sda).thenElse(srcover, dstover);
     return alphas.zeroColors() + colors.zeroAlphas();
 }
 #undef XFERMODE
 
 // Some xfermodes use math like divide or sqrt that's best done in floats 1 pixel at a time.
-#define XFERMODE(Name) static SkPMFloat SK_VECTORCALL Name(SkPMFloat d, SkPMFloat s)
+#define XFERMODE(Name) static SkPMFloat SK_VECTORCALL Name(SkPMFloat s, SkPMFloat d)
 
 XFERMODE(ColorDodge) {
     auto sa = s.alphas(),
          da = d.alphas(),
          isa = Sk4f(1)-sa,
          ida = Sk4f(1)-da;
 
     auto srcover = s + d*isa,
          dstover = d + s*ida,
          otherwise = sa * Sk4f::Min(da, (d*sa)*(sa-s).approxInvert()) + s*ida + d*isa;
@@ skipping 43 matching lines @@
     auto alpha  = s + d*isa;
     auto colors = s*ida + d*isa + (s2 <= sa).thenElse(darkSrc, liteSrc);           // Case 1 or 2/3?
 
     return alpha * SkPMFloat(1,0,0,0) + colors * SkPMFloat(0,1,1,1);
 }
 #undef XFERMODE
 
 // A reasonable fallback mode for doing AA is to simply apply the transfermode first,
 // then linearly interpolate the AA.
 template <Sk4px (SK_VECTORCALL *Mode)(Sk4px, Sk4px)>
-static Sk4px SK_VECTORCALL xfer_aa(Sk4px d, Sk4px s, Sk4px aa) {
-    Sk4px bw = Mode(d, s);
+static Sk4px SK_VECTORCALL xfer_aa(Sk4px s, Sk4px d, Sk4px aa) {
+    Sk4px bw = Mode(s, d);
     return (bw * aa + d * aa.inv()).div255();
 }
 
 // For some transfermodes we specialize AA, either for correctness or performance.
 #define XFERMODE_AA(Name) \
-    template <> Sk4px SK_VECTORCALL xfer_aa<Name>(Sk4px d, Sk4px s, Sk4px aa)
+    template <> Sk4px SK_VECTORCALL xfer_aa<Name>(Sk4px s, Sk4px d, Sk4px aa)
 
 // Plus' clamp needs to happen after AA.  skia:3852
 XFERMODE_AA(Plus) {  // [ clamp( (1-AA)D + (AA)(S+D) ) == clamp(D + AA*S) ]
     return d.saturatedAdd(s.approxMulDiv255(aa));
 }
 
 #undef XFERMODE_AA
 
 class Sk4pxXfermode : public SkProcCoeffXfermode {
 public:
     typedef Sk4px (SK_VECTORCALL *Proc4)(Sk4px, Sk4px);
     typedef Sk4px (SK_VECTORCALL *AAProc4)(Sk4px, Sk4px, Sk4px);
 
     Sk4pxXfermode(const ProcCoeff& rec, SkXfermode::Mode mode, Proc4 proc4, AAProc4 aaproc4)
         : INHERITED(rec, mode)
         , fProc4(proc4)
         , fAAProc4(aaproc4) {}
 
     void xfer32(SkPMColor dst[], const SkPMColor src[], int n, const SkAlpha aa[]) const override {
         if (NULL == aa) {
-            Sk4px::MapDstSrc(n, dst, src, fProc4);
+            Sk4px::MapDstSrc(n, dst, src, [&](const Sk4px& dst4, const Sk4px& src4) {
+                return fProc4(src4, dst4);
+            });
         } else {
-            Sk4px::MapDstSrcAlpha(n, dst, src, aa, fAAProc4);
+            Sk4px::MapDstSrcAlpha(n, dst, src, aa,
+                    [&](const Sk4px& dst4, const Sk4px& src4, const Sk4px& alpha) {
+                return fAAProc4(src4, dst4, alpha);
+            });
         }
     }
 
     void xfer16(uint16_t dst[], const SkPMColor src[], int n, const SkAlpha aa[]) const override {
         if (NULL == aa) {
-            Sk4px::MapDstSrc(n, dst, src, fProc4);
+            Sk4px::MapDstSrc(n, dst, src, [&](const Sk4px& dst4, const Sk4px& src4) {
+                return fProc4(src4, dst4);
+            });
         } else {
-            Sk4px::MapDstSrcAlpha(n, dst, src, aa, fAAProc4);
+            Sk4px::MapDstSrcAlpha(n, dst, src, aa,
+                    [&](const Sk4px& dst4, const Sk4px& src4, const Sk4px& alpha) {
+                return fAAProc4(src4, dst4, alpha);
+            });
         }
     }
 
 private:
     Proc4 fProc4;
     AAProc4 fAAProc4;
     typedef SkProcCoeffXfermode INHERITED;
 };
 
 class SkPMFloatXfermode : public SkProcCoeffXfermode {
@@ skipping 14 matching lines @@
         for (int i = 0; i < n; i++) {
             SkPMColor dst32 = SkPixel16ToPixel32(dst[i]);
             dst32 = aa ? this->xfer32(dst32, src[i], aa[i])
                        : this->xfer32(dst32, src[i]);
             dst[i] = SkPixel32ToPixel16(dst32);
         }
     }
 
 private:
     inline SkPMColor xfer32(SkPMColor dst, SkPMColor src) const {
-        return fProcF(SkPMFloat(dst), SkPMFloat(src)).round();
+        return fProcF(SkPMFloat(src), SkPMFloat(dst)).round();
     }
 
     inline SkPMColor xfer32(SkPMColor dst, SkPMColor src, SkAlpha aa) const {
         SkPMFloat s(src),
                   d(dst),
-                  b(fProcF(d,s));
+                  b(fProcF(s,d));
         // We do aa in full float precision before going back down to bytes, because we can!
         SkPMFloat a = Sk4f(aa) * Sk4f(1.0f/255);
         b = b*a + d*(Sk4f(1)-a);
         return b.round();
     }
 
     ProcF fProcF;
     typedef SkProcCoeffXfermode INHERITED;
 };
 
@@ skipping 33 matching lines @@
     #undef CASE
 
         default: break;
     }
     return nullptr;
 }
 
 } // namespace SK_NS_OPTS
 
 #endif//Sk4pxXfermode_DEFINED