X64: Reenabled RSet.

src/x64/assembler-x64.h

Index: src/x64/assembler-x64.h
diff --git a/src/x64/assembler-x64.h b/src/x64/assembler-x64.h
index ad4721d6ad55c428f703a905cc636a5d6f5a9a16..db58fc355c2d31e73a66d10371823e76ecc9c5a1 100644
--- a/src/x64/assembler-x64.h
+++ b/src/x64/assembler-x64.h
@@ -44,15 +44,25 @@ namespace internal {
 
 // Test whether a 64-bit value is in a specific range.
 static inline bool is_uint32(int64_t x) {
-  const int64_t kUInt32Mask = V8_INT64_C(0xffffffff);
+  static const int64_t kUInt32Mask = V8_INT64_C(0xffffffff);
   return x == (x & kUInt32Mask);
 }
 
 static inline bool is_int32(int64_t x) {
-  const int64_t kMinIntValue = V8_INT64_C(-0x80000000);
+  static const int64_t kMinIntValue = V8_INT64_C(-0x80000000);
   return is_uint32(x - kMinIntValue);
 }
 
+static inline bool is_int32(uint64_t x) {

[William Hesse, 2009/08/03 09:03:28]

I can forsee a bug happening because of this function.  I think we sometimes use
a uint64_t for data that really includes negative values.  The proper function
for a person to use in that case is is_int32(int64_t x), but they would need a
cast to use it.  Could this be called uint_is_int32 to make it clear to the user
that it should be used carefully?

[Lasse Reichstein, 2009/08/03 10:45:45]

I'm not sure we shouldn't find a better way to bite someone in the behind for
holding signed data in unsigned variables, and using them without casting.
But let's rename for now. The function isn't used yet, it was just added for
symmetry.

+  static const uint64_t kMaxIntValue = V8_UINT64_C(0x80000000);
+  return x < kMaxIntValue;
+}
+
+static inline bool is_uint32(uint64_t x) {
+  static const uint64_t kMaxUIntValue = V8_UINT64_C(0x100000000);
+  return x < kMaxUIntValue;
+}
+
 // CPU Registers.
 //
 // 1) We would prefer to use an enum, but enum values are assignment-