FYI, there are tests that may be useful at
Are there any performance comparisons for this work? I'd love to see a blog post informing about the ongoing changes and what this means e. g. for DOM performance (with up-to-date numbers). Though this can wait until the majority of Bugs (esp. for document Obj Bindings) have been fixed.
A followup blog post to
would be indeed nice.
We definitely have performance numbers. See http://dromaeo.com/?id=186075,186085,186086,186076 for example, and more to come as we convert actual DOM nodes and documents and such...
There are also the obvious wins on ImageData performance (bug 746773 comment 14) and whatnot, and on WebGL performance.
Since the bindings are landing piecemeal and the JS engine is improving in parallel, disentangling the two effects can be difficult, though. I can't tell you which one is responsible for the changes in the Dromaeo dom-query numbers above, for example. So do you want to see numbers comparing to Fx13 (before any of this stuff landed) or some other peformance comparison?
This is indeed helpful. It would be nice to have kind of a summary after all the ParisBindings work is done.
Comparing Fx 13 with any current version will probably not tell you very much, as we now have a totally new JIT engine (though comparing Fx 13 to any pre-IM version may help, as well as any comparisons between different IM version).
What about having something like AWFY (or actually extend AWFY with Dromaeo)? That may also help with finding regressions.
> as we now have a totally new JIT engine
Well. Part of the bindings work was closer integration with the new JIT that's allowed by the new bindings; in fact that's one of the major sources of speedup. So for classes we convert now comparing before/after behavior makes sense, but for classes we converted before fx18 there was a noticeable speedup from Ion landing.
> or actually extend AWFY with Dromaeo
AWFY runs in a JS shell, so does not have DOM access, last I checked. Something like AWFY for DOM stuff would be great, of course. ;)