Open Bug 915195 Opened 12 years ago Updated 3 years ago

InvalidateFrameSubtree and the stuff under it could be faster

Categories

(Core :: Layout, defect)

x86_64
Linux
defect

Tracking

()

People

(Reporter: smaug, Unassigned)

Details

(Keywords: perf)

Last night I wondered why FF seemed to take quite some cpu power occasionally, so I got few profiles and nsIFrame::InvalidateFrameSubtree and stuff under it seemed to take significant amout time. InvalidateDirectRenderingObservers is also in the profiles. There are few simple minor optimizations; for example make sure nsSVGEffects::InvalidateDirectRenderingObservers isn't even called if there are no observers (so, inline parts of nsSVGEffects::InvalidateDirectRenderingObservers), The profiler also suggested making InvalidateFrameSubtree non-recursive. I wonder if bug 870430 would have helped.
Ah, looks like we have something like bug 870430 in the tree now.
No longer depends on: 870430
Do you have a link to a profile and the URL of the page you were viewing when you saw the perf issues? Also which code are you referring to in comment 1?
Keywords: perf
part of the profile, https://pastebin.mozilla.org/3026585, total time%, and self time%
This could be related to bug 912769
Severity: normal → S3
You need to log in before you can comment on or make changes to this bug.