1106100 - Figure out why jQuery's Data.accepts is ending up with MGetPropertyCache, not MGetDOMProperty for .nodeType

Status: NEW

(Core :: JavaScript Engine: JIT, defect, P5)

Description

[Boris Zbarsky [:bzbarsky]]

My bet is that jQuery itself passes a non-DOM object through here, and then we can't do the faster path, but worth checking.

Comment 1

[Andrew Overholt [:overholt]]

Boris, is this still relevant? Is the potential performance win here beneficial enough that we'd tell jQuery people about it?

Comment 2

[Boris Zbarsky [:bzbarsky]]

This came up in profiles in bug 1104898.  I don't know whether it's still an issue with modern jQuery or how big the win would be in a non-synthetic testcase...  Needs measuring.

Comment 3

[Tom S [:evilpie]]

I fixed some cases in CacheIR where "nodeType" was being looked up on primitive types like strings, and we are still missing an IC for looking up nodeType on arguments objects. I didn't look to deeply into where this happens, but I think this was jQuery or some related library. So I think some places where they use nodeType are quite polymorphic.

Comment 4

[Boris Zbarsky [:bzbarsky]]

Right, see comment 0.  In that case, it really might be worth it mentioning this to them....

Comment 5

[Desigan Chinniah [:cyberdees] [:dees] [London - GMT]]

(In reply to Boris Zbarsky [:bz] (still a bit busy) (if a patch has no decent message, automatic r-) from comment #4)
> Right, see comment 0.  In that case, it really might be worth it mentioning
> this to them....

Boris - Could you use our ML to bring this to their attention, please?

Comment 6

[Boris Zbarsky [:bzbarsky]]

Well, the first question is whether it's still an issue with modern jQuery as opposed to the ancient ones the benchmarks use.  Someone needs to check that.