[Webkit-unassigned] [Bug 149713] DFG peephole does not have effect on CompareEq / CompareStrictEq because of MovHint

bugzilla-daemon at webkit.org bugzilla-daemon at webkit.org
Thu Oct 1 10:23:49 PDT 2015


https://bugs.webkit.org/show_bug.cgi?id=149713

--- Comment #2 from Yusuke Suzuki <utatane.tea at gmail.com> ---
(In reply to comment #1)
> When and if we implement a fix that allows the peephole detection to "skip"
> the MovHint, we should check how much of a speed-up it is.
> 
> If it is not a speed-up, then instead of writing a patch that extends the
> peephole to skip MovHint, we should write a patch that removes all code for
> those peepholes that are currently dead because of the MovHint.  That
> peephole code is a pain to maintain and it's possible that it buys us
> nothing.

Right. This peep hole optimization is easily broken because the assumption the optimization rely on is fragile.
If it does not provide significant performance improvement, dropping these code is better.

-- 
You are receiving this mail because:
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.webkit.org/pipermail/webkit-unassigned/attachments/20151001/fbdec821/attachment.html>


More information about the webkit-unassigned mailing list