[Webkit-unassigned] [Bug 17510] Acid3 test 26 takes >33ms

bugzilla-daemon at webkit.org bugzilla-daemon at webkit.org
Mon May 26 22:37:53 PDT 2008


------- Comment #41 from mjs at apple.com  2008-05-26 22:37 PDT -------
I took a fresh profile and looked at a top-down view to get a rough estimate of
how readily we can improve times on this test. We need to cut out approximately
23% of the current execution time (a 1.3x speedup).

Top-down, I see the following mutually exclusive categories of time:

7.5% self time in KJS::Machine::privateExecute (could potentially be reduced by
optimizing JS)
16.8% total time in JSNode::insertBefore
8.5% total time in JSNode::removeChild
8.4% total time in JSObject::defaultValue
7.7% total time in jsString()
4.8% total time in JSElement::setAttribute
4.0% total time in jsDocumentPrototypeFunctionCreateElement
3.1% total time in stringProtoFuncMatch
3.1% total time in JSNode::getValueProperty
2.3% total time in JSNode::appendChild
2.0% total time in jsDocumentPrototypeFunctionCreateTextNode
1.6% total time in makeFunction
1.6% total time in UString concatenation constructor
1.5% total time in jsDocumentPrototypeFunctionGetElementById
1.3% total time in DateObjectImp::construct
1.1% total time in JSCharacterData::getValueProperty

Seems like optimizing string concatenation could plausibly remove half the
jsString() time and half the UString constructor time so that is a plausible
big win. Devirtualizing defaultValue could also be a good payoff. Reducing the
insertBefore or removeChild time seems like a challenge though may be necessary
to win.

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

More information about the webkit-unassigned mailing list