1506033 - 4.77 - 13.35% displaylist_mutate (linux64-qr, windows10-64-qr) regression on push c41ac37391713511296521e6dfecd54739a870f7 (Thu Nov 8 2018)

Status: VERIFIED FIXED

(Core :: Graphics: WebRender, defect, P2)

Description

[Ionuț Goldan [:igoldan]]

Talos has detected a Firefox performance regression from push:

https://hg.mozilla.org/integration/autoland/pushloghtml?changeset=c41ac37391713511296521e6dfecd54739a870f7

As author of one of the patches included in that push, we need your help to address this regression.

Regressions:

 13%  displaylist_mutate windows10-64-qr opt e10s stylo     4,920.73 -> 5,577.63
  5%  displaylist_mutate linux64-qr opt e10s stylo          5,082.01 -> 5,324.61


You can find links to graphs and comparison views for each of the above tests at: https://treeherder.mozilla.org/perf.html#/alerts?id=17430

On the page above you can see an alert for each affected platform as well as a link to a graph showing the history of scores for this test. There is also a link to a treeherder page showing the Talos jobs in a pushlog format.

To learn more about the regressing test(s), please see: https://wiki.mozilla.org/Performance_sheriffing/Talos/Tests

For information on reproducing and debugging the regression, either on try or locally, see: https://wiki.mozilla.org/Performance_sheriffing/Talos/Running

*** Please let us know your plans within 3 business days, or the offending patch(es) will be backed out! ***

Our wiki page outlines the common responses and expectations: https://wiki.mozilla.org/Performance_sheriffing/Talos/RegressionBugsHandling

Comment 1

[Ionuț Goldan [:igoldan]]

Will soon post the Gecko profiles.

Comment 2

[Ionuț Goldan [:igoldan]]

Here are the Gecko profiles for displaylist_mutate on Windows 10 QR:

before: https://perf-html.io/from-url/https%3A%2F%2Fqueue.taskcluster.net%2Fv1%2Ftask%2FaWVg5YnzTT656W0UfgM7wg%2Fruns%2F0%2Fartifacts%2Fpublic%2Ftest_info%2Fprofile_displaylist_mutate.zip

after: https://perf-html.io/from-url/https%3A%2F%2Fqueue.taskcluster.net%2Fv1%2Ftask%2FEFGFl9BDQ-unQ9J02VwJ9w%2Fruns%2F0%2Fartifacts%2Fpublic%2Ftest_info%2Fprofile_displaylist_mutate.zip


and for displaylist_mutate on Linux 64bit QR:

before: https://perf-html.io/from-url/https%3A%2F%2Fqueue.taskcluster.net%2Fv1%2Ftask%2Fe6HBDF6ETN66OyWzkL8ReQ%2Fruns%2F0%2Fartifacts%2Fpublic%2Ftest_info%2Fprofile_displaylist_mutate.zip

after: https://perf-html.io/from-url/https%3A%2F%2Fqueue.taskcluster.net%2Fv1%2Ftask%2FJU53MoXITcGNvHZrkZ0XUw%2Fruns%2F0%2Fartifacts%2Fpublic%2Ftest_info%2Fprofile_displaylist_mutate.zip

Comment 3

[Bobby Holley (:bholley)]

This is from https://github.com/servo/webrender/pull/3270

The 13% regression on windows seems big, not sure if it was expected.

Comment 4

[Glenn Watson [:gw]]

That's definitely unexpected from this patch. I'll investigate on Monday.

Comment 5

[Glenn Watson [:gw]]

I profiled this and it's the same problem as the last regression - this test is very sensitive to the size of the PrimitiveInstance structure, because of the number that are created and how WR creates and iterates them.

I'm planning to land most of the picture caching work in the next 1-2 weeks. A side effect of this work is that the size of the PrimitiveInstance structure should be significantly reduced (hopefully by 50% or more), and we will iterate them more efficiently (only when they have been moved by their positioning node relative to rasterization root, rather than every instance every frame).

I'll take this bug, and plan to have significant performance improvements such that we can close it in a couple of weeks.

Comment 6

[Glenn Watson [:gw]]

I think this can be closed, now that https://bugzilla.mozilla.org/show_bug.cgi?id=1507257 has brought the dl_mutate time down to ~4700?

Comment 7

[Ionuț Goldan [:igoldan]]

(In reply to Glenn Watson [:gw] from comment #6)
> I think this can be closed, now that
> https://bugzilla.mozilla.org/show_bug.cgi?id=1507257 has brought the
> dl_mutate time down to ~4700?

Yes, I totally agree. Actually, the performance is even better than the results prior to the regression! Thanks a lot!

Comment 8

[Ionuț Goldan [:igoldan]]

I believe bug 1509302 did the fix. I'll link it here.