Use generated Neon version of MemCopy() on ARM, if platform supports it.

src/platform.h

Index: src/platform.h
===================================================================
--- src/platform.h	(revision 14076)
+++ src/platform.h	(working copy)
@@ -113,6 +113,10 @@
 typedef intptr_t AtomicWord;
 #endif
 
+#if defined(V8_TARGET_ARCH_IA32) || defined(V8_TARGET_ARCH_ARM)
+#define V8_GENERATED_MEMCPY 1
+#endif
+
 class Semaphore;
 class Mutex;
 
@@ -336,7 +340,7 @@
 
   static void ReleaseStore(volatile AtomicWord* ptr, AtomicWord value);
 
-#if defined(V8_TARGET_ARCH_IA32)
+#if defined(V8_GENERATED_MEMCPY)
   // Copy memory area to disjoint memory area.
   static void MemCopy(void* dest, const void* src, size_t size);
   // Limit below which the extra overhead of the MemCopy function is likely
@@ -344,12 +348,12 @@
   static const int kMinComplexMemCopy = 64;
   typedef void (*MemCopyFunction)(void* dest, const void* src, size_t size);
 
-#else  // V8_TARGET_ARCH_IA32
+#else  // V8_GENERATED_MEMCPY
   static void MemCopy(void* dest, const void* src, size_t size) {
     memcpy(dest, src, size);
   }
   static const int kMinComplexMemCopy = 16 * kPointerSize;
-#endif  // V8_TARGET_ARCH_IA32
+#endif  // V8_GENERATED_MEMCPY
 
   static int GetCurrentProcessId();