Support for MIPS in architecture independent files....

src/globals.h

 // Copyright 2006-2009 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 28 matching lines @@
 #define V8_HOST_ARCH_X64 1
 #define V8_HOST_ARCH_64_BIT 1
 #define V8_HOST_CAN_READ_UNALIGNED 1
 #elif defined(_M_IX86) || defined(__i386__)
 #define V8_HOST_ARCH_IA32 1
 #define V8_HOST_ARCH_32_BIT 1
 #define V8_HOST_CAN_READ_UNALIGNED 1
 #elif defined(__ARMEL__)
 #define V8_HOST_ARCH_ARM 1
 #define V8_HOST_ARCH_32_BIT 1
+#elif defined(_MIPS_ARCH_MIPS32R2)
+#define V8_HOST_ARCH_MIPS 1
+#define V8_HOST_ARCH_32_BIT 1
 #else
 #error Your host architecture was not detected as supported by v8
 #endif
 
 #if defined(V8_TARGET_ARCH_X64) || defined(V8_TARGET_ARCH_IA32)
 #define V8_TARGET_CAN_READ_UNALIGNED 1
 #elif V8_TARGET_ARCH_ARM
+#elif V8_TARGET_ARCH_MIPS
+#define V8_TARGET_CAN_READ_UNALIGNED 0

[Søren Thygesen Gjesse, 2010/01/19 22:59:12]

defining V8_TARGET_CAN_READ_UNALIGNED to 0 should not be needed.

[Alexandre, 2010/01/22 23:08:42]

Removed.
On 2010/01/19 22:59:12, Søren Gjesse wrote:

 #else
 #error Your target architecture is not supported by v8
 #endif
 
 // Support for alternative bool type. This is only enabled if the code is
 // compiled with USE_MYBOOL defined. This catches some nasty type bugs.
 // For instance, 'bool b = "false";' results in b == true! This is a hidden
 // source of bugs.
 // However, redefining the bool type does have some negative impact on some
 // platforms. It gives rise to compiler warnings (i.e. with
@@ skipping 96 matching lines @@
 // Should be a recognizable hex value tagged as a heap object pointer.
 #ifdef V8_HOST_ARCH_64_BIT
 const Address kZapValue =
     reinterpret_cast<Address>(V8_UINT64_C(0xdeadbeedbeadbeed));
 const Address kHandleZapValue =
     reinterpret_cast<Address>(V8_UINT64_C(0x1baddead0baddead));
 const Address kFromSpaceZapValue =
     reinterpret_cast<Address>(V8_UINT64_C(0x1beefdad0beefdad));
 #else
 const Address kZapValue = reinterpret_cast<Address>(0xdeadbeed);
+#ifndef V8_TARGET_ARCH_MIPS
 const Address kHandleZapValue = reinterpret_cast<Address>(0xbaddead);
+#else
+// On mips 0xbaddead is the encoding of jump to 0xeb77ab4.

[Søren Thygesen Gjesse, 2010/01/19 22:59:12]

I don't think this is important. The handle zap value is placed where object
pointers used to be, and will most likely be used as an object value and not
executed code. Getting a segfault should be fine.

[Alexandre, 2010/01/22 23:08:42]

I wrote this in the early development stage. It can safely be removed.
On 2010/01/19 22:59:12, Søren Gjesse wrote:

+// This can be misleading when debugging and developping as we get a segfault
+// instead of an illegal instruction when executing it.
+// The 0b111011 opcode is reserved for future use. Executing this instruction
+// raises a Reserved Instruction Exception error on MIPS not using this opcode.
+const Address kHandleZapValue = reinterpret_cast<Address>(0xec000000);
+#endif
 const Address kFromSpaceZapValue = reinterpret_cast<Address>(0xbeefdad);
 #endif
 
 
 // Constants relevant to double precision floating point numbers.
 
 // Quiet NaNs have bits 51 to 62 set, possibly the sign bit, and no
 // other bits set.
 const uint64_t kQuietNaNMask = static_cast<uint64_t>(0xfff) << 51;
 // If looking only at the top 32 bits, the QNaN mask is bits 19 to 30.
@@ skipping 387 matching lines @@
 
   Dest dest;
   memcpy(&dest, &source, sizeof(dest));
   return dest;
 }
 
 
 } }  // namespace v8::internal
 
 #endif  // V8_GLOBALS_H_