Open
Bug 1228879
Opened 9 years ago
Updated 3 years ago
spend lots of time in nsRegion::Or called from ContainerState::FinishPaintedLayerData, with lots of polygons in a leaflet map
Categories
(Core :: Web Painting, defect)
Core
Web Painting
Tracking
()
NEW
Tracking | Status | |
---|---|---|
firefox45 | --- | affected |
People
(Reporter: dbaron, Unassigned)
References
Details
(Keywords: perf)
I was profiling the leaflet-osm-conversion branch of https://github.com/dbaron/timezone-map to figure out why it was slow in Gecko. One thing I noticed is that we're spending significant amounts of time (>10%) in nsRegion::Or called from ContainerState::FinishPaintedLayerData.
This appears to be to union into containingPaintedLayerData->mDispatchToContentHitRegion and containingPaintedLayerData->mMaybeHitRegion, although it might also be including the calls through nsLayoutUtils::TransformToAncestorAndCombineRegions.
Comment 1•9 years ago
|
||
We should maybe be simplifying these regions, but it's not as straightforward as the other places that we simplify regions.
We don't want to expand mHitRegion (otherwise the compositor will falsely report a hit) so maybe would could SimplifyInward the hit region, and put area that SimplifyInward discards into the maybe hit region or dispatch to content region.
The maybe hit region and dispatch to content region could be simplified outward, but we are trading off having to ask the main thread about taps for a simpler region here.
Assignee | ||
Updated•7 years ago
|
Component: Layout: View Rendering → Layout: Web Painting
Updated•4 years ago
|
![]() |
||
Updated•4 years ago
|
Webcompat Priority: --- → ?
Updated•3 years ago
|
Webcompat Priority: ? → ---
Comment 2•3 years ago
|
||
ContainerState::FinishPaintedLayerData and related code no longer exists. We should compare/profile this with current code and see if there is any issue.
Updated•3 years ago
|
Severity: normal → S3
You need to log in
before you can comment on or make changes to this bug.
Description
•