2001628 - [Linux] Remove compositor pause/resume and use async rendering

Status: RESOLVED FIXED

(Core :: Widget: Gtk, enhancement, P3)

Description

[Martin Stránský [:stransky] (ni? me)]

Investigate removal of synced compositor updates (Bug 1997782):

https://searchfox.org/firefox-main/rev/70425199e9a5fa80aa7419fa51763013a67226e1/widget/gtk/nsWindow.cpp#10051
https://searchfox.org/firefox-main/rev/70425199e9a5fa80aa7419fa51763013a67226e1/widget/gtk/nsWindow.cpp#6929

Comment 1

[Martin Stránský [:stransky] (ni? me)]

https://bugzilla.mozilla.org/show_bug.cgi?id=1997782#c9 is relevant here.

Comment 2

[Martin Stránský [:stransky] (ni? me)]

Attachment: WIP: Bug 2001628 [Linux] Don't restart compositor after nsWindow::Create()

Comment 3

[Martin Stránský [:stransky] (ni? me)]

Almost done bud I don't understand why we need to call remoteRenderer->SendResume(); (unfortunately synced) after mCompositorWidgetDelegate set.

I'd expect the copmpositor is already running as CompositorInitiallyPaused() is set on Android only:
https://searchfox.org/firefox-main/rev/1c9d86edc5d91b67ba8c858a053de40e1b98dc95/widget/nsIWidget.cpp#1495

Comment 4

[Martin Stránský [:stransky] (ni? me)]

SendResumeAsync(); may be a workaround if we really need to start compositor (but why?).

Comment 5

[Martin Stránský [:stransky] (ni? me)]

Looks like we should set the rendering surface to surface provider at nsWindow::Create() and then create EGLSurface at RenderCompositorEGL::RenderCompositorEGL(). We also need to make sure GLX rendering is not broken.

Comment 6

[Martin Stránský [:stransky] (ni? me)]

Attachment: Bug 2001628 [Linux] Make mWindowSurface persistent r?emilio

Comment 7

[Martin Stránský [:stransky] (ni? me)]

try: https://treeherder.mozilla.org/jobs?repo=try&landoCommitID=170043

Comment 9

[Pulsebot]

Pushed by [e-mail]:
https://github.com/mozilla-firefox/firefox/commit/ac30d4d7974f
https://hg.mozilla.org/integration/autoland/rev/7abccda1bbb1
[Linux] Make mWindowSurface persistent r=emilio

Comment 10

[Pulsebot]

Pushed by [e-mail]:
https://github.com/mozilla-firefox/firefox/commit/6e1be6241390
https://hg.mozilla.org/integration/autoland/rev/0ee39346fe18
Revert "Bug 2001628 [Linux] Make mWindowSurface persistent r=emilio" for causing bc failures @ browser_bookmark_context_menu_contents.js

Comment 11

[amarc]

Backed out for causing bc failures

• Backout link
• Push with failures
• Failure Log
• Failure Log 2

Comment 12

[Martin Stránský [:stransky] (ni? me)]

I can reproduce the timeout locally but I have no idea what causes it:

 0:13.21 GECKO(201597) [Parent 201597: Main Thread]: D/Widget [7fa76d7d3900]: nsWindow::OnShellConfigureEvent() [960, 556] -> [800 x 500] scale 2.00 (scaled size 1600 x 1000)
 0:13.21 GECKO(201597) [Parent 201597: Main Thread]: D/Widget [7fa76d7d3900]: OnPropertyNotifyEvent(_NET_FRAME_EXTENTS)
 0:13.21 GECKO(201597) [Parent 201597: Main Thread]: D/Widget [7fa76d7d3900]: nsWindow::OnShellConfigureEvent() [960, 556] -> [1 x 1] scale 2.00 (scaled size 2 x 2)

The library toplevel window is rendered 1x1 size after _NET_FRAME_EXTENTS.

Comment 13

[Martin Stránský [:stransky] (ni? me)]

Fixed by latest patch update. Let's wait when code freeze is over.

Comment 14

[Pulsebot]

Pushed by [e-mail]:
https://github.com/mozilla-firefox/firefox/commit/7eeebeb3ecfc
https://hg.mozilla.org/integration/autoland/rev/4898d067ee28
[Linux] Make mWindowSurface persistent r=emilio

Comment 15

[amarc]

https://hg.mozilla.org/mozilla-central/rev/4898d067ee28

Comment 16

[Florin Bilt]

(In reply to amarc from comment #15)

https://hg.mozilla.org/mozilla-central/rev/4898d067ee28

Perfherder has detected a talos performance change from push 4898d067ee28f052f9d1c57ffde37dc0aad68e33.

No action is required from the author; this comment is provided for informational purposes only.

Improvement	Test	Platform	Options	Absolute values [old vs new]
8%	twinopen ext+twinopen:twinopen.html (doc)	linux1804-64-shippable-qr	e10s fission stylo webrender-sw	156.74 ms -> 144.76 ms

Need Help or Information?

If you have any questions, please reach out to [e-mail]. Alternatively, you can find help on Slack by joining #perf-help, and on Matrix you can find help by joining #perftest.

Details of the alert can be found in the alert summary, including links to graphs and comparisons for each of the affected tests.