We forgot to fix the DebugBreakIC... to know that the registers contained
GCable
values. Fixed in http://codereview.chromium.org/660257
http://codereview.chromium.org/608031/diff/6009/6014
File src/ia32/ic-ia32.cc (right):
http://codereview.chromium.org/608031/diff/6009/6014#newcode817
src/ia32/ic-ia32.cc:817: __ fstp_s(Operand(edi, ebx, times_4, 0));
Not valid. ret(4) pops 4 bytes _after_ popping the return address.
On 2010/02/25 11:53:52, fschneider wrote:
Tiny optimization: To remove the pop-instruction instead of
__ push(ecx)
__ fild_s(...)
__ pop(ecx)
__ fstp_s(...)
__ ret(0)
you could just do
__ push(ecx)
__ fild_s(...)
__ fstp_s(...)
__ ret(4)
And hoist __ ret(0) into the switch-statement for the other cases.
http://codereview.chromium.org/608031
--
v8-dev mailing list
[email protected]
http://groups.google.com/group/v8-dev
