[Webkit-unassigned] [Bug 67993] Prediction tracking is not precise enough

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=67993





--- Comment #10 from Filip Pizlo <fpizlo at apple.com>  2011-09-14 11:56:21 PST ---
(In reply to comment #8)
> (From update of attachment 107272 [details])
> What happens if we see "o.a" and o varies between an array and an ordinary object?  do we get strong-object or weak-array?

If by "ordinary object" you mean JSFinalObject then we get Strong|PredictArray|PredictFinalObject.  If by "ordinary object" you mean some subclass of JSObject other than JSFinalObject then we get Strong|PredictArray|PredictObjectOther.  If by "ordinary object" you mean that we inferred that by seeing that there were GetById/GetByVal's executed on the object, or that it came out of some op that's known to return objects but we don't know anything else about it, then we just get Strong|PredictArray.  If by "ordinary object" you mean that we saw a JSFinalObjects and other subclasses of JSObject then we get Strong|PredictArray|PredictFinalObject|PredictObjectOther.  Note that there are bits left behind for other kinds of objects in the future (functions, typed arrays, etc), so this is still different from PredictObject.  Currently we won't ever get PredictObject in any PredictedType variable because of those missing bits, and because nothing in the analysis every returns PredictO
 bject.

-- 
Configure bugmail: https://bugs.webkit.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.