Xfermode: SSE2 implementation of overlay_modeproc

src/opts/SkColor_opts_SSE2.h

Index: src/opts/SkColor_opts_SSE2.h
diff --git a/src/opts/SkColor_opts_SSE2.h b/src/opts/SkColor_opts_SSE2.h
index 960c48a023126ee304f0d0505d1b09054c542b46..3a4baf27d995905e2db12fe8378badc7e9f5d3b7 100644
--- a/src/opts/SkColor_opts_SSE2.h
+++ b/src/opts/SkColor_opts_SSE2.h
@@ -10,6 +10,14 @@
 
 #include <emmintrin.h>
 
+static inline  __m128i Multiply32_SSE2(const __m128i& a, const __m128i& b) {
+    __m128i r1 = _mm_mul_epu32(a, b);
+    __m128i r2 = _mm_mul_epu32(_mm_srli_si128(a, 4), _mm_srli_si128(b, 4));

[mtklein, 2014/04/22 15:18:21]

Can you narrate this method a bit more, at least with a comment at the top about
what math it does exactly on a and b?  I only ask because Multiply32_SSE2 is a
pretty generic name, and I can't find any corresponding non-SIMD code to compare
it with.

[qiankun, 2014/04/23 07:40:05]

Done.

+    __m128i r = _mm_unpacklo_epi32(_mm_shuffle_epi32(r1, _MM_SHUFFLE(0,0,2,0)),
+                                   _mm_shuffle_epi32(r2, _MM_SHUFFLE(0,0,2,0)));
+    return r;
+}
+
 static inline __m128i SkAlpha255To256_SSE2(const __m128i& alpha) {
     return _mm_add_epi32(alpha, _mm_set1_epi32(1));
 }