Improve SkARGB32_A8_BlitMask_SSE2

src/opts/SkBlitRow_opts_SSE2.cpp

Index: src/opts/SkBlitRow_opts_SSE2.cpp
diff --git a/src/opts/SkBlitRow_opts_SSE2.cpp b/src/opts/SkBlitRow_opts_SSE2.cpp
index 391b24c8673809c9a40a0225be52129af2369237..363cdab9f0390daf7847dba10f95cbeebfdda199 100644
--- a/src/opts/SkBlitRow_opts_SSE2.cpp
+++ b/src/opts/SkBlitRow_opts_SSE2.cpp
@@ -436,21 +436,20 @@ void SkARGB32_A8_BlitMask_SSE2(void* device, size_t dstRB, const void* maskPtr,
             __m128i c_256 = _mm_set1_epi16(256);
             __m128i c_1 = _mm_set1_epi16(1);
             __m128i src_pixel = _mm_set1_epi32(color);
             while (count >= 4) {
                 // Load 4 pixels each of src and dest.
                 __m128i dst_pixel = _mm_load_si128(d);
 
                 //set the aphla value
-                __m128i src_scale_wide =  _mm_set_epi8(0, *(mask+3),\
-                                0, *(mask+3),0, \
-                                *(mask+2),0, *(mask+2),\
-                                0,*(mask+1), 0,*(mask+1),\
-                                0, *mask,0,*mask);
+                __m128i src_scale_wide = _mm_cvtsi32_si128(*reinterpret_cast<const uint32_t*>(mask));
+                src_scale_wide = _mm_unpacklo_epi8(src_scale_wide,
+                                                   _mm_setzero_si128());
+                src_scale_wide = _mm_unpacklo_epi16(src_scale_wide, src_scale_wide);
 
                 //call SkAlpha255To256()
                 src_scale_wide = _mm_add_epi16(src_scale_wide, c_1);
 
                 // Get red and blue pixels into lower byte of each word.
                 __m128i dst_rb = _mm_and_si128(rb_mask, dst_pixel);
                 __m128i src_rb = _mm_and_si128(rb_mask, src_pixel);