Add sk_memcpy32 to Memcpy bench.

bench/MemcpyBench.cpp

Index: bench/MemcpyBench.cpp
diff --git a/bench/MemcpyBench.cpp b/bench/MemcpyBench.cpp
index aec2a4774d9efc45f85fc2c32b37425d561952f5..3fd6dcbcfb56492f6acea9ed548d3f2bdde62006 100644
--- a/bench/MemcpyBench.cpp
+++ b/bench/MemcpyBench.cpp
@@ -8,6 +8,7 @@
 #include "SkBenchmark.h"
 #include "SkRandom.h"
 #include "SkTemplates.h"
+#include "SkUtils.h"
 
 template <typename Memcpy32>
 class Memcpy32Bench : public SkBenchmark {
@@ -147,6 +148,13 @@ BENCH(memcpy32_sse2_unalign, 1000)
 BENCH(memcpy32_sse2_unalign, 10000)
 BENCH(memcpy32_sse2_unalign, 100000)
 
+// Test our chosen best, from SkUtils.h
+BENCH(sk_memcpy32, 10)
+BENCH(sk_memcpy32, 100)
+BENCH(sk_memcpy32, 1000)
+BENCH(sk_memcpy32, 10000)
+BENCH(sk_memcpy32, 100000)
+
 #endif // SK_CPU_SSE_LEVEL >= SK_CPU_SSE_LEVEL_SSE2
 
 #undef BENCH