Minor bug in int32 TypeRecordingBinaryOpStub.

BUG=none
TEST=none
Patch by Rodolph Perfetta from ARM Ltd.

Review URL: http://codereview.chromium.org/6612017


git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@7055 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
This commit is contained in:
sgjesse@chromium.org 2011-03-04 07:36:52 +00:00
parent 88b60ebc24
commit 18a9bf69c5

View File

@ -872,12 +872,11 @@ void FloatingPointHelper::DoubleIs32BitInteger(MacroAssembler* masm,
// Exponent greater than 31 cannot yield 32-bit integers.
// Also, a positive value with an exponent equal to 31 is outside of the
// signed 32-bit integer range.
__ tst(src1, Operand(HeapNumber::kSignMask));
__ cmp(scratch, Operand(30), eq); // Executed for positive. If exponent is 30
// the gt condition will be "correct" and
// the next instruction will be skipped.
__ cmp(scratch, Operand(31), ne); // Executed for negative and positive where
// exponent is not 30.
// Another way to put it is that if (exponent - signbit) > 30 then the
// number cannot be represented as an int32.
Register tmp = dst;
__ sub(tmp, scratch, Operand(src1, LSR, 31));
__ cmp(tmp, Operand(30));
__ b(gt, not_int32);
// - Bits [21:0] in the mantissa are not null.
__ tst(src2, Operand(0x3fffff));