ARM: Implement memcpy using NEON.

src/platform.h

 // Copyright 2012 the V8 project authors. All rights reserved.
 // Redistribution and use in source and binary forms, with or without
 // modification, are permitted provided that the following conditions are
 // met:
 //
 //     * Redistributions of source code must retain the above copyright
 //       notice, this list of conditions and the following disclaimer.
 //     * Redistributions in binary form must reproduce the above
 //       copyright notice, this list of conditions and the following
 //       disclaimer in the documentation and/or other materials provided
@@ skipping 297 matching lines @@
   // Maximum size of the virtual memory.  0 means there is no artificial
   // limit.
   static intptr_t MaxVirtualMemory();
 
   // Returns the double constant NAN
   static double nan_value();
 
   // Support runtime detection of Cpu implementer
   static CpuImplementer GetCpuImplementer();
 
+  // Support runtime detection of Cpu implementer
+  static CpuPart GetCpuPart(CpuImplementer implementer);
+
   // Support runtime detection of VFP3 on ARM CPUs.
   static bool ArmCpuHasFeature(CpuFeature feature);
 
   // Support runtime detection of whether the hard float option of the
   // EABI is used.
   static bool ArmUsingHardFloat();
 
   // Support runtime detection of FPU on MIPS CPUs.
   static bool MipsCpuHasFeature(CpuFeature feature);
 
   // Returns the activation frame alignment constraint or zero if
   // the platform doesn't care. Guaranteed to be a power of two.
   static int ActivationFrameAlignment();
 
 #if defined(V8_TARGET_ARCH_IA32)
   // Limit below which the extra overhead of the MemCopy function is likely
   // to outweigh the benefits of faster copying.
   static const int kMinComplexMemCopy = 64;
 
   // Copy memory area. No restrictions.
   static void MemMove(void* dest, const void* src, size_t size);
   typedef void (*MemMoveFunction)(void* dest, const void* src, size_t size);
 
   // Keep the distinction of "move" vs. "copy" for the benefit of other
   // architectures.
   static void MemCopy(void* dest, const void* src, size_t size) {
     MemMove(dest, src, size);
   }
-#else  // V8_TARGET_ARCH_IA32
+#elif defined(V8_HOST_ARCH_ARM)
+  typedef void (*MemCopyUint8Function)(uint8_t* dest,
+                                       const uint8_t* src,
+                                       size_t size);
+  static MemCopyUint8Function memcopy_uint8_function;
+  static void MemCopyUint8Wrapper(uint8_t* dest,
+                                  const uint8_t* src,
+                                  size_t chars) {
+    memcpy(dest, src, chars);
+  }
+  // For values < 16, the assembler function is slower than the inlined C code.
+  static const int kMinComplexMemCopy = 16;
+  static void MemCopy(void* dest, const void* src, size_t size) {
+    (*memcopy_uint8_function)(reinterpret_cast<uint8_t*>(dest),
+                              reinterpret_cast<const uint8_t*>(src),
+                              size);
+  }
+  static void MemMove(void* dest, const void* src, size_t size) {
+    memmove(dest, src, size);
+  }
+
+  typedef void (*MemCopyUint16Uint8Function)(uint16_t* dest,
+                                             const uint8_t* src,
+                                             size_t size);
+  static MemCopyUint16Uint8Function memcopy_uint16_uint8_function;
+  static void MemCopyUint16Uint8Wrapper(uint16_t* dest,
+                                        const uint8_t* src,
+                                        size_t chars);
+  // For values < 12, the assembler function is slower than the inlined C code.
+  static const int kMinComplexConvertMemCopy = 12;
+  static void MemCopyUint16Uint8(uint16_t* dest,
+                                 const uint8_t* src,
+                                 size_t size) {
+    (*memcopy_uint16_uint8_function)(dest, src, size);
+  }
+#else
   // Copy memory area to disjoint memory area.
   static void MemCopy(void* dest, const void* src, size_t size) {
     memcpy(dest, src, size);
   }
   static void MemMove(void* dest, const void* src, size_t size) {
     memmove(dest, src, size);
   }
   static const int kMinComplexMemCopy = 16 * kPointerSize;
 #endif  // V8_TARGET_ARCH_IA32
 
@@ skipping 369 matching lines @@
   static uint16_t HToN(uint16_t value);
   static uint16_t NToH(uint16_t value);
   static uint32_t HToN(uint32_t value);
   static uint32_t NToH(uint32_t value);
 };
 
 
 } }  // namespace v8::internal
 
 #endif  // V8_PLATFORM_H_