Committed as revision 979.

http://codereview.chromium.org/13344/diff/208/12
File src/virtual-frame-ia32.cc (right):

http://codereview.chromium.org/13344/diff/208/12#newcode504
Line 504: !target.reg().is(source.reg())) {
On 2008/12/14 13:05:04, Kevin Millikin wrote:
> As we discussed, the source and targets could be the same register but
with
> different sync states.  In that case, the source's sync bit may need
to be
> cleared, or a move generated to set the target's sync bit.

Done.

http://codereview.chromium.org/13344/diff/208/12#newcode526
Line 526: SyncElementAt(i);
On 2008/12/14 13:05:04, Kevin Millikin wrote:
> It seems better to call sync (which will work on the source register)
before
> changing the reference counts to reflect the transfer.

Done.

http://codereview.chromium.org/13344/diff/208/12#newcode538
Line 538: end = last_move_blocked;
On 2008/12/12 18:13:03, iposva wrote:
> This is potentially a premature optimization. These stacks should
never be very
> large to start with.

You are right.  But should we actually remove it?  The locals could be
big.  Some people are passing big arrays as argument arrays.  Are they
on the frame?  Not removed, for now.

http://codereview.chromium.org/13344

--~--~---------~--~----~------------~-------~--~----~
v8-dev mailing list
[email protected]
http://groups.google.com/group/v8-dev
-~----------~----~----~----~------~----~------~--~---

Reply via email to