3-15% speedup to HardLight / Overlay xfermodes.

src/opts/SkNx_sse.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 SkNx_sse_DEFINED
 #define SkNx_sse_DEFINED
 
@@ skipping 239 matching lines @@
 
     static SkNi Min(const SkNi& a, const SkNi& b) {
         // No unsigned _mm_min_epu16, so we'll shift into a space where we can use the
         // signed version, _mm_min_epi16, then shift back.
         const uint16_t top = 0x8000; // Keep this separate from _mm_set1_epi16 or MSVC will whine.
         const __m128i top_8x = _mm_set1_epi16(top);
         return _mm_add_epi8(top_8x, _mm_min_epi16(_mm_sub_epi8(a.fVec, top_8x),
                                                   _mm_sub_epi8(b.fVec, top_8x)));
     }
 
+    SkNi thenElse(const SkNi& t, const SkNi& e) const {
+        return _mm_or_si128(_mm_and_si128   (fVec, t.fVec),
+                            _mm_andnot_si128(fVec, e.fVec));
+    }
+
     template <int k> uint16_t kth() const {
         SkASSERT(0 <= k && k < 8);
         return _mm_extract_epi16(fVec, k);
     }
 
     __m128i fVec;
 };
 
 template <>
 class SkNi<16, uint8_t> {
@@ skipping 34 matching lines @@
         return _mm_or_si128(_mm_and_si128   (fVec, t.fVec),
                             _mm_andnot_si128(fVec, e.fVec));
     }
 
     __m128i fVec;
 };
 
 }  // namespace
 
 #endif//SkNx_sse_DEFINED