Closed Bug 788093 Opened 12 years ago Closed 8 years ago

OpenLayers.js map with a vector layer on which features are visible is terribly slow

Categories

(Core :: SVG, defect)

x86
All
defect
Not set
normal

Tracking

()

RESOLVED WORKSFORME

People

(Reporter: nirvn.asia, Unassigned)

References

()

Details

(Keywords: perf, Whiteboard: [invalidation] [in-the-wild] [external-report])

Go to http://licadho-cambodia.org/live/ and play with the slider for the map to add / remove concessions. As more concessions are added on the map, dragging the map around becomes significantly slower. With all concessions added to the map, it takes around half a second to render map when dragging around. This issue is not present under chromium, where the map dragging process remains fluid independently of the number of concessions.
Changing platform to all as I could confirm bug on a windows 7 machine. I'm also moving this to the SVG component as the polygon features layer is SVG rendered.
Component: Graphics → SVG
OS: Linux → All
Hmm, this isn't only affecting dragging of map, it's actually affecting scrolling of page when map is visible.
... which is odd as the scrolling slowdown occurs even when the map region (contained within a div, which includes the svg layer) remains entirely visible (I'd have assumed that if the div / svg region remains entirely visible, gecko would not need to redraw).
Status: UNCONFIRMED → NEW
Ever confirmed: true
Keywords: perf
Summary: openlayers map with a vector layer on which features are visible is terribly slow (in comparison to chromium) on linux → OpenLayers.js map with a vector layer on which features are visible is terribly slow
SPS Profile against Mozilla/5.0 (Windows NT 6.1; WOW64; rv:18.0) Gecko/18.0 Firefox/18.0 ID:20120909030609 with HWA on: http://people.mozilla.com/~bgirard/cleopatra/?report=e8740327b194d24ff3bf9124b0e1742418fcbeb0
another site where this issue is visible: http://flooddemo.estil-jennyl.com/ -- author suggests viewing site with chrome :o/
Is there a way I can help profiling this issue? Firefox is turning out to be the worse performer in this context (including ie9, which I've tested yesterday).
Reuters published the above map on its anti-corruption blog: http://www.trust.org/trustlaw/blogs/anti-corruption-views/the-great-cambodian-giveaway
I can't really reproduce this on Mac. I would note that we seem to be invalidating the entire browser content area while dragging the map though, which sucks. XtC4UaLL - your profile doesn't seem to show anything obvious. Did you maximize the number of concessions being shown, stop the profiler, restart it, immediately pan the map around for a few seconds, then Analyze the profile? It would be great if you could do that to get as many samples in the bottleneck code as possible.
Jonathan, that's weird. My computer is pretty high-end and I can still see a noticeable difference when dragging between chromium and firefox. Both on windows and linux. NVIDIA graphic card.
(In reply to Jonathan Watt [:jwatt] from comment #8) I don't remember, but I can see two Perf related Issues (compared to Chrome) here: * moving the Slider and the Painting of the Concessions (even without panning the Map) SPS Profile: http://people.mozilla.com/~bgirard/cleopatra/?report=4824e8d7f958793e8cc6bdbb69bc9bc7494b234a * panning the Map with all available Concessions shown. There's clearly a Difference compared to panning without any Concessions shown SPS Profile: http://people.mozilla.com/~bgirard/cleopatra/?report=c73f63a05800b3dcc0cdb1f9a84d3beb48770f62 For both Issues I didn't zoom in. I'm on a Box with an AMD 4850 Card with D2D/D3D10 & Direct2D Azure Backend on.
Whiteboard: [jwatt:invalidation]
Whiteboard: [jwatt:invalidation] → [jwatt:invalidation] [in-the-wild] [external-report]
Paint flashing doesn't show so much invalidation nowadays. Anyone on Mac/Windows care to test what Firefox's performance is like compared to other browsers you care about?
Whiteboard: [jwatt:invalidation] [in-the-wild] [external-report] → [invalidation] [in-the-wild] [external-report]
Maybe Daniel, if nobody else is still active on this bug.
Flags: needinfo?(dholbert)
(In reply to Jonathan Watt [:jwatt] from comment #11) > Anyone on Mac/Windows care to test what Firefox's performance is like compared to > other browsers you care about? (I suspect you meant Linux/Windows not Mac/Windows, since you're on Mac yourself & this was originally filed as a Linux bug.) When I drag the slider on Linux, the repaints seems smooth to me. On Windows, it's slightly janky (small noticeable delay), though Chrome feels equivalently janky on Windows.
Flags: needinfo?(dholbert)
Thanks, Daniel. And yes, I meant Linux/Windows. :)
Status: NEW → RESOLVED
Closed: 8 years ago
Resolution: --- → WORKSFORME
I also have this problem of performance on my computer, tiles loading is very very slow while map navigation. In Chromium all is ok, fluid and fast. All is on my computer as standalone (apache, tile maps). Firefox 54.0.1 Linux Debian 9 (stretch) OpenLayers 4.2.
... it seems that if disabling µBlock origin/ghostery it works much better, not as fast as Chromium (with µBlock origin/ghostery), but much better !
You need to log in before you can comment on or make changes to this bug.