922048 - Peacekeeper domJQueryBasicFilters slower than in Chrome

Status: VERIFIED FIXED

(Core :: JavaScript Engine, defect)

Description

[Jan de Mooij [:jandem]]

Attachment: Testcase

Original test here: http://www.peacekeeper.therichins.net/test19.html

Attaching a standalone version with a non-minified version of jQuery (1.6.4)

Chrome 31:  1671 ms
Nightly 27: 2842 ms

Looks like we're spending time in the Yarr interpreter and various VM calls, so filing this as a JS bug for now. Will investigate more after filing some other bugs.

Comment 1

[Jan de Mooij [:jandem]]

Oh the test just does:

    $(".frontPageNewsItem:first");
    $(".frontPageNewsItem:last");
    $(".frontPageNewsItem div:not(.title)");
    $(".frontPageNewsItem:even");
    $(".frontPageNewsItem:eq(3)");
    $("body *:header");

All the Peacekeeper jQuery tests are testing selectors :(

Comment 2

[Boris Zbarsky [:bzbarsky]]

That's what most jQuery tests test.  Note bug 786160 and whatever djvj is working on nowadays for dromaeo-css, which is similar stuff.

Comment 3

[Jan de Mooij [:jandem]]

(In reply to Jan de Mooij [:jandem] from comment #0)
> Chrome 31:  1671 ms
> Nightly 27: 2842 ms

With the fixes for bug 723640, bug 890722 and probably bug 863966, this is now down to 2056 ms.

Comment 4

[Jan de Mooij [:jandem]]

(In reply to Jan de Mooij [:jandem] from comment #3)
> (In reply to Jan de Mooij [:jandem] from comment #0)
> > Chrome 31:  1671 ms
> > Nightly 27: 2842 ms
> 
> With the fixes for bug 723640, bug 890722 and probably bug 863966, this is
> now down to 2056 ms.

With bug 928423 we're at ~1950 ms so we closed most of the gap. Unfortunately we spend about 200-300 ms in the YARR interpreter (bug 929507). We can probably win a bit more by optimizing the DOM error reporting code (see discussion in bug 929950).

Comment 5

[Olli Pettay [:smaug][[e-mail]]]

On 64bit linux I get now 1900-1975ms in opt m-i build and 1850-1925 in 31 chrome.
A few days old opt m-c is ~2150ms

Comment 6

[Jan de Mooij [:jandem]]

Looks like bug 933475 was a small win here (1656 ms -> 1588 ms) and we're slightly faster than Chrome now (1619 ms).

Comment 7

[Olli Pettay [:smaug][[e-mail]]]

Yeah, I get faster values too, on linux.
Bug 932837 will make things even better.

But I'd say this bug as such is fixed.

Comment 8

[Ioana (away)]

Tested with Firefox 28.0b3, older buggy Firefox, Chrome on Mac OS X 10.7.5, Windows 7 64bit and Ubuntu 12.10 64bit:

Windows averages - Chrome 1557ms, buggy build 3624ms, Beta 1730ms
Ubuntu averages - Chrome 1680ms, buggy build 35380ms, Beta 1543ms
Mac averages - Chrome 3203ms, buggy build 3285ms, Beta 3165ms