Replace SSE optimization of Color32A_D565

src/opts/opts_check_x86.cpp

 /*
  * Copyright 2009 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 "SkBitmapFilter_opts_SSE2.h"
 #include "SkBitmapProcState_opts_SSE2.h"
 #include "SkBitmapProcState_opts_SSSE3.h"
@@ skipping 198 matching lines @@
 };
 
 SkBlitRow::Proc16 SkBlitRow::PlatformFactory565(unsigned flags) {
     if (supports_simd(SK_CPU_SSE_LEVEL_SSE2)) {
         return platform_16_procs[flags];
     } else {
         return NULL;
     }
 }
 
-static const SkBlitRow::ColorProc16 platform_565_colorprocs_SSE4[] = {
-    Color32A_D565_SSE4,                 // Color32A_D565,
+static const SkBlitRow::ColorProc16 platform_565_colorprocs_SSE2[] = {
+    Color32A_D565_SSE2,                 // Color32A_D565,
     NULL,                               // Color32A_D565_Dither
 };
 
 SkBlitRow::ColorProc16 SkBlitRow::PlatformColorFactory565(unsigned flags) {
-    if (supports_simd(SK_CPU_SSE_LEVEL_SSE41)) {
-        return platform_565_colorprocs_SSE4[flags];
+/* If you're thinking about writing an SSE4 version of this, do check it's
+ * actually faster on Atom. Our original SSE4 version was slower than this
+ * SSE2 version on Silvermont, and only marginally faster on a Core i7,
+ * mainly due to the MULLD timings.
+ */
+    if (supports_simd(SK_CPU_SSE_LEVEL_SSE2)) {
+        return platform_565_colorprocs_SSE2[flags];
     } else {
         return NULL;
     }
 }
 
 static const SkBlitRow::Proc32 platform_32_procs_SSE2[] = {
     NULL,                               // S32_Opaque,
     S32_Blend_BlitRow32_SSE2,           // S32_Blend,
     S32A_Opaque_BlitRow32_SSE2,         // S32A_Opaque
     S32A_Blend_BlitRow32_SSE2,          // S32A_Blend,
@@ skipping 169 matching lines @@
     } else {
         return SkPlatformXfermodeFactory_impl(rec, mode);
     }
 }
 
 SkXfermodeProc SkPlatformXfermodeProcFactory(SkXfermode::Mode mode);
 
 SkXfermodeProc SkPlatformXfermodeProcFactory(SkXfermode::Mode mode) {
     return NULL;
 }