Add a note about (x*y+255)/256.

src/core/Sk4px.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 Sk4px_DEFINED
 #define Sk4px_DEFINED
 
@@ skipping 85 matching lines @@
     // These just keep the types as Sk4px so the user doesn't have to keep casting.
     Sk4px operator + (const Sk4px& o) const { return INHERITED::operator+(o); }
     Sk4px operator - (const Sk4px& o) const { return INHERITED::operator-(o); }
     Sk4px operator < (const Sk4px& o) const { return INHERITED::operator<(o); }
     Sk4px thenElse(const Sk4px& t, const Sk4px& e) const { return INHERITED::thenElse(t,e); }
 
     // Generally faster than (*this * o).div255().
     // May be incorrect by +-1, but is always exactly correct when *this or o is 0 or 255.
     Sk4px approxMulDiv255(const Sk16b& o) const {
         // (x*y + x) / 256 meets these criteria.  (As of course does (x*y + y) / 256 by symmetry.)
+        // FYI: (x*y + 255) / 256 also meets these criteria.  In my brief testing, it was slower.
         return this->widenLo().addNarrowHi(*this * o);
     }
 
     // A generic driver that maps fn over a src array into a dst array.
     // fn should take an Sk4px (4 src pixels) and return an Sk4px (4 dst pixels).
     template <typename Fn, typename Dst>
     static void MapSrc(int n, Dst* dst, const SkPMColor* src, const Fn& fn) {
         SkASSERT(dst);
         SkASSERT(src);
         // This looks a bit odd, but it helps loop-invariant hoisting across different calls to fn.
@@ skipping 127 matching lines @@
     #if SK_CPU_SSE_LEVEL >= SK_CPU_SSE_LEVEL_SSE2
         #include "../opts/Sk4px_SSE2.h"
     #elif defined(SK_ARM_HAS_NEON)
         #include "../opts/Sk4px_NEON.h"
     #else
         #include "../opts/Sk4px_none.h"
     #endif
 #endif
 
 #endif//Sk4px_DEFINED