Summary: | DFG double voting may be overzealous in the case of variables that end up being used as integers | ||||||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
Product: | WebKit | Reporter: | Filip Pizlo <fpizlo> | ||||||||||||
Component: | JavaScriptCore | Assignee: | Nobody <webkit-unassigned> | ||||||||||||
Status: | RESOLVED FIXED | ||||||||||||||
Severity: | Normal | CC: | ggaren | ||||||||||||
Priority: | P2 | ||||||||||||||
Version: | 528+ (Nightly build) | ||||||||||||||
Hardware: | All | ||||||||||||||
OS: | All | ||||||||||||||
Attachments: |
|
Description
Filip Pizlo
2012-03-22 20:21:31 PDT
Created attachment 133423 [details]
work in progress
Still working on it but at least it compiles now.
Created attachment 133431 [details]
the patch
Comment on attachment 133431 [details] the patch View in context: https://bugs.webkit.org/attachment.cgi?id=133431&action=review r+ with the one fix. > Source/JavaScriptCore/dfg/DFGPredictionPropagationPhase.cpp:199 > + mergeDefaultFlags(node); Need to or this into changed. Created attachment 133581 [details]
the patch
Created attachment 133611 [details]
the patch
Created attachment 133615 [details]
the patch
Removed the EscapesAsInt portion of the patch because it was causing perf regressions, and the case where it was beneficial was sort of contrived to begin with.
> and the case where it was beneficial was sort of contrived to begin with.
... Interesting. FWIW, kind of struck me that way, too.
(In reply to comment #7) > > and the case where it was beneficial was sort of contrived to begin with. > > ... Interesting. FWIW, kind of struck me that way, too. Yeah Gavin had to convince me to trust the numbers on this one. ;-) Landed in http://trac.webkit.org/changeset/112015 |