Fix memcpy32_sse2_unalign.

bench/MemcpyBench.cpp

 /*
  * Copyright 2014 Google Inc.
  *
  * Use of this source code is governed by a BSD-style license that can be
  * found in the LICENSE file.
  */
 
 #include "SkBenchmark.h"
 #include "SkRandom.h"
 #include "SkTemplates.h"
@@ skipping 115 matching lines @@
         __m128i a = _mm_loadu_si128(src128++);
         __m128i b = _mm_loadu_si128(src128++);
         __m128i c = _mm_loadu_si128(src128++);
         __m128i d = _mm_loadu_si128(src128++);
 
         _mm_storeu_si128(dst128++, a);
         _mm_storeu_si128(dst128++, b);
         _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)
 BENCH(memcpy32_sse2_unalign, 100000)
 
 #endif // SK_CPU_SSE_LEVEL >= SK_CPU_SSE_LEVEL_SSE2
 
 #undef BENCH