Fix memcpy32_sse2_unalign.

bench/MemcpyBench.cpp

Index: bench/MemcpyBench.cpp
diff --git a/bench/MemcpyBench.cpp b/bench/MemcpyBench.cpp
index 0bb7b617b46dbf10ee8fa8581f9192ceaa2cca64..2bc03d005445bb18195983202efecfc03788caf3 100644
--- a/bench/MemcpyBench.cpp
+++ b/bench/MemcpyBench.cpp
@@ -133,17 +133,16 @@ static void memcpy32_sse2_unalign(uint32_t* dst, const uint32_t* src, int count)
         _mm_storeu_si128(dst128++, c);
         _mm_storeu_si128(dst128++, d);
 
+        dst += 16;
+        src += 16;

[f(malita), 2014/05/22 16:46:34]

Is the compiler smart enough to hoist these?

[mtklein, 2014/05/22 16:48:59]

Probably.  I was worried that _I_ wasn't smart enough to hoist them correctly. 
Would it have been,

dst += 16 * (count / 16);
src += 16 * (count / 16);

?

[mtklein, 2014/05/22 16:59:02]

Done.

While I was looking at it, I realized that, if I'm right, then the _align
version was half-broken too: src128 isn't aligned (dst128 is).  Made the same
change there.

         count -= 16;
     }
 
-    dst = reinterpret_cast<uint32_t*>(dst128);
-    src = reinterpret_cast<const uint32_t*>(src128);
     while (count --> 0) {
         *dst++ = *src++;
     }
 }
-// skia:2589: Crashing on ChromeOS Alex bot.  TODO(mtklein): why?
-//BENCH(memcpy32_sse2_unalign, 10)
+BENCH(memcpy32_sse2_unalign, 10)
 BENCH(memcpy32_sse2_unalign, 100)
 BENCH(memcpy32_sse2_unalign, 1000)
 BENCH(memcpy32_sse2_unalign, 10000)