Convert Color32 code to perfect blend.

src/opts/SkBlitRow_opts_SSE2.cpp

 /*
  * Copyright 2012 The Android Open Source Project
  *
  * Use of this source code is governed by a BSD-style license that can be
  * found in the LICENSE file.
  */
 
 #include <emmintrin.h>
 #include "SkBitmapProcState_opts_SSE2.h"
 #include "SkBlitRow_opts_SSE2.h"
@@ skipping 216 matching lines @@
     while (count > 0) {
         *dst = SkBlendARGB32(*src, *dst, alpha);
         src++;
         dst++;
         count--;
     }
 }
 
 #define SK_SUPPORT_LEGACY_COLOR32_MATHx
 
-/* SSE2 version of Color32()
- * portable version is in core/SkBlitRow_D32.cpp
- */
-// Color32 and its SIMD specializations use the blend_256_round_alt algorithm
-// from tests/BlendTest.cpp.  It's not quite perfect, but it's never wrong in the
-// interesting edge cases, and it's quite a bit faster than blend_perfect.
-//
-// blend_256_round_alt is our currently blessed algorithm.  Please use it or an analogous one.
+/* SSE2 version of Color32(), portable version is in core/SkBlitRow_D32.cpp */
+// Color32 and its SIMD specializations use the blend_perfect algorithm from tests/BlendTest.cpp.
+// An acceptable alternative is blend_256_round_alt, which is faster but not quite perfect.
 void Color32_SSE2(SkPMColor dst[], const SkPMColor src[], int count, SkPMColor color) {
     switch (SkGetPackedA32(color)) {
         case   0: memmove(dst, src, count * sizeof(SkPMColor)); return;
         case 255: sk_memset32(dst, color, count);               return;
     }
 
-    __m128i colorHigh = _mm_unpacklo_epi8(_mm_setzero_si128(), _mm_set1_epi32(color));
-#ifdef SK_SUPPORT_LEGACY_COLOR32_MATH  // blend_256_plus1_trunc, busted
-    __m128i colorAndRound = colorHigh;
-#else                          // blend_256_round_alt, good
-    __m128i colorAndRound = _mm_add_epi16(colorHigh, _mm_set1_epi16(128));
-#endif
-
-    unsigned invA = 255 - SkGetPackedA32(color);
-#ifdef SK_SUPPORT_LEGACY_COLOR32_MATH  // blend_256_plus1_trunc, busted
-    __m128i invA16 = _mm_set1_epi16(invA);
-#else                          // blend_256_round_alt, good
-    SkASSERT(invA + (invA >> 7) < 256);  // We should still fit in the low byte here.
-    __m128i invA16 = _mm_set1_epi16(invA + (invA >> 7));
-#endif
+    __m128i color_2x_high = _mm_unpacklo_epi8(_mm_setzero_si128(), _mm_set1_epi32(color)),
+             invA_8x      = _mm_set1_epi16(255 - SkGetPackedA32(color));
 
     // Does the core work of blending color onto 4 pixels, returning the resulting 4 pixels.
-    auto kernel = [&](const __m128i& src4) -> __m128i {
-        __m128i lo = _mm_mullo_epi16(invA16, _mm_unpacklo_epi8(src4, _mm_setzero_si128())),
-                hi = _mm_mullo_epi16(invA16, _mm_unpackhi_epi8(src4, _mm_setzero_si128()));
-        return _mm_packus_epi16(_mm_srli_epi16(_mm_add_epi16(colorAndRound, lo), 8),
-                                _mm_srli_epi16(_mm_add_epi16(colorAndRound, hi), 8));
+    auto kernel = [&](const __m128i& src_4x) -> __m128i {
+        __m128i lo = _mm_mullo_epi16(invA_8x, _mm_unpacklo_epi8(src_4x, _mm_setzero_si128())),
+                hi = _mm_mullo_epi16(invA_8x, _mm_unpackhi_epi8(src_4x, _mm_setzero_si128()));
+    #ifndef SK_SUPPORT_LEGACY_COLOR32_MATH
+        lo = _mm_add_epi16(lo, _mm_set1_epi16(128));
+        hi = _mm_add_epi16(hi, _mm_set1_epi16(128));
+        lo = _mm_add_epi16(lo, _mm_srli_epi16(lo, 8));
+        hi = _mm_add_epi16(hi, _mm_srli_epi16(hi, 8));
+    #endif
+        return _mm_packus_epi16(_mm_srli_epi16(_mm_add_epi16(color_2x_high, lo), 8),
+                                _mm_srli_epi16(_mm_add_epi16(color_2x_high, hi), 8));
     };
 
     while (count >= 8) {
         __m128i dst0 = kernel(_mm_loadu_si128((const __m128i*)(src+0))),
                 dst4 = kernel(_mm_loadu_si128((const __m128i*)(src+4)));
         _mm_storeu_si128((__m128i*)(dst+0), dst0);
         _mm_storeu_si128((__m128i*)(dst+4), dst4);
         src   += 8;
         dst   += 8;
         count -= 8;
@@ skipping 941 matching lines @@
                 uint32_t dst_expanded = SkExpand_rgb_16(*dst);
                 dst_expanded = dst_expanded * (SkAlpha255To256(255 - a) >> 3);
                 // now src and dst expanded are in g:11 r:10 x:1 b:10
                 *dst = SkCompact_rgb_16((src_expanded + dst_expanded) >> 5);
             }
             dst += 1;
             DITHER_INC_X(x);
         } while (--count != 0);
     }
 }