1501172 - 2.76 - 3.3% displaylist_mutate (linux64-qr, windows10-64-qr) regression on push 2d8a1d405333036b4adce643e4893f2da29957be (Sat Oct 20 2018)

Status: RESOLVED DUPLICATE of bug 1426770

(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=2d8a1d405333036b4adce643e4893f2da29957be

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

Regressions:

  3%  displaylist_mutate windows10-64-qr opt e10s stylo     4,815.45 -> 4,974.33
  3%  displaylist_mutate linux64-qr opt e10s stylo          4,745.64 -> 4,876.74


You can find links to graphs and comparison views for each of the above tests at: https://treeherder.mozilla.org/perf.html#/alerts?id=17017

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/Buildbot/Talos/Tests

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

Comment 1

[Kartikaya Gupta (email:[e-mail])]

That WR update just contained servo/webrender#3218

Comment 2

[Glenn Watson [:gw]]

That patch has no functional changes - it changes our memory layout for display items. I believe this test case has a huge number of display items, in which case it's possible that just the cache / memory behavior may have regressed?

I don't expect any noticeable regression on real world pages with smaller numbers of display items. As this is part of a large set of incremental changes (picture caching) that are expected to significantly improve performance, I think it's fine to accept this test case regression in the interim.

Comment 3

[Bobby Holley (:bholley)]

I think Matt's familiar with this test and has done some work to optimize it. Matt, does that sound reasonable to you?

Comment 4

[Matt Woodrow (:mattwoodrow)]

Yeah it does. We're still quite a bit slower on this test, and a few percent worse won't make much difference.

I plan on doing another round of profiling for this (once I'm sick of looking at SVGs), so we can probably make this back up then.

Comment 5

[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%2FfhEIMVpfSt-A0pRXZB2JHg%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%2FCultYIRlSDGbNZbHu5YCuQ%2Fruns%2F0%2Fartifacts%2Fpublic%2Ftest_info%2Fprofile_displaylist_mutate.zip

Comment 6

[Ionuț Goldan [:igoldan]]

Any updates here?

Comment 7

[Jeff Muizelaar [:jrmuizel]]

We're tracking display mutate performance in bug 1426770. We're expecting movement up and down on this test for the next little while.