Add SSE4 check to BlurImage optimization.

src/opts/SkBlurImage_opts_SSE2.cpp

 /*
  * Copyright 2013 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 "SkBitmap.h"
 #include "SkBlurImage_opts_SSE2.h"
@@ skipping 37 matching lines @@
         __m128i sum = zero;
         const SkPMColor* p = src;
         for (int i = 0; i < rightBorder; ++i) {
             sum = _mm_add_epi32(sum, expand(*p));
             p += srcStrideX;
         }
 
         const SkPMColor* sptr = src;
         SkColor* dptr = dst;
         for (int x = 0; x < width; ++x) {
-#if 0
+#if SK_CPU_SSE_LEVEL >= SK_CPU_SSE_LEVEL_SSE42

[mtklein, 2014/05/14 17:15:56]

I think the comment below is correct, right?  Only needs SK_CPU_SSE_LEVEL_SSE41?

(My reference is https://software.intel.com/sites/landingpage/IntrinsicsGuide)

[henrik.smiding, 2014/05/15 15:12:05]

Done.

             // In SSE4.1, this would be
             __m128i result = _mm_mullo_epi32(sum, scale);
 #else
             // But SSE2 has no PMULLUD, so we must do AG and RB separately.
             __m128i tmp1 = _mm_mul_epu32(sum, scale);
             __m128i tmp2 = _mm_mul_epu32(_mm_srli_si128(sum, 4),
                                          _mm_srli_si128(scale, 4));
             __m128i result = _mm_unpacklo_epi32(_mm_shuffle_epi32(tmp1, _MM_SHUFFLE(0,0,2,0)),
                                                 _mm_shuffle_epi32(tmp2, _MM_SHUFFLE(0,0,2,0)));
 #endif
@@ skipping 34 matching lines @@
 bool SkBoxBlurGetPlatformProcs_SSE2(SkBoxBlurProc* boxBlurX,
                                     SkBoxBlurProc* boxBlurY,
                                     SkBoxBlurProc* boxBlurXY,
                                     SkBoxBlurProc* boxBlurYX) {
     *boxBlurX = SkBoxBlur_SSE2<kX, kX>;
     *boxBlurY = SkBoxBlur_SSE2<kY, kY>;
     *boxBlurXY = SkBoxBlur_SSE2<kX, kY>;
     *boxBlurYX = SkBoxBlur_SSE2<kY, kX>;
     return true;
 }