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

src/arm/assembler-arm.h

Index: src/arm/assembler-arm.h
===================================================================
--- src/arm/assembler-arm.h	(revision 14076)
+++ src/arm/assembler-arm.h	(working copy)
@@ -440,7 +440,22 @@
   p15 = 15
 };
 
+enum NeonElementIndex {
+  element_0 = 0,
+  element_1 = 1,
+  element_2 = 2,
+  element_3 = 3,
+  element_4 = 4,
+  element_5 = 5,
+  element_6 = 6,
+  element_7 = 7
+};
 
+enum NeonWritebackType {
+  Writeback = 0,
+  NoWriteback = 1
+};
+
 // -----------------------------------------------------------------------------
 // Machine instruction Operands
 
@@ -895,6 +910,10 @@
   void ldm(BlockAddrMode am, Register base, RegList dst, Condition cond = al);
   void stm(BlockAddrMode am, Register base, RegList src, Condition cond = al);
 
+  // Prefetch instructions.
+  void pld(Register base, int offset = 0);
+  void pldw(Register base, int offset = 0);
+
   // Exception-generating instructions and debugging support
   void stop(const char* msg,
             Condition cond = al,
@@ -1098,6 +1117,56 @@
              const DwVfpRegister src,
              const Condition cond = al);
 
+  // NEON instructions
+  void vld1(int size,
+            const Register base,
+            const DwVfpRegister first,
+            const DwVfpRegister last,
+            NeonWritebackType writeback = NoWriteback,
+            int align_bytes = 1);
+  void vld1(int size,
+            const Register base,
+            const DwVfpRegister first,
+            NeonElementIndex element_index,
+            NeonWritebackType writeback = NoWriteback,
+            int align_bytes = 1);
+  void vld4(int size,
+            const Register src,
+            const DwVfpRegister first,
+            const DwVfpRegister last,
+            NeonWritebackType writeback = NoWriteback,
+            int align_bytes = 1);
+  void vld4(int size,
+            const Register src,
+            const DwVfpRegister first,
+            NeonElementIndex element_index,
+            NeonWritebackType writeback = NoWriteback,
+            int align_bytes = 1);
+  void vst1(int size,
+            const Register base,
+            const DwVfpRegister first,
+            const DwVfpRegister last,
+            NeonWritebackType writeback = NoWriteback,
+            int align_bytes = 1);
+  void vst1(int size,
+            const Register base,
+            const DwVfpRegister first,
+            NeonElementIndex element_index,
+            NeonWritebackType writeback = NoWriteback,
+            int align_bytes = 1);
+  void vst4(int size,
+            const Register src,
+            const DwVfpRegister first,
+            const DwVfpRegister last,
+            NeonWritebackType writeback = NoWriteback,
+            int align_bytes = 1);
+  void vst4(int size,
+            const Register src,
+            const DwVfpRegister first,
+            NeonElementIndex element_index,
+            NeonWritebackType writeback = NoWriteback,
+            int align_bytes = 1);
+
   // Pseudo instructions
 
   // Different nop operations are used by the code generator to detect certain