1459847 - 4.64 - 6.59% displaylist_mutate (linux64) regression on push f8876cbee5fa17c374115ed04404a3a3fc61ba40 (Mon May 7 2018)

Status: RESOLVED WONTFIX

(Core :: XUL, defect)

Description

[Ionuț Goldan [:igoldan]]

Talos has detected a Firefox performance regression from push:

https://hg.mozilla.org/integration/autoland/pushloghtml?changeset=f8876cbee5fa17c374115ed04404a3a3fc61ba40

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

Regressions:

  7%  displaylist_mutate linux64 opt e10s stylo     3,341.83 -> 3,562.17
  5%  displaylist_mutate linux64 pgo e10s stylo     2,999.51 -> 3,138.60


You can find links to graphs and comparison views for each of the above tests at: https://treeherder.mozilla.org/perf.html#/alerts?id=13100

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

[Ionuț Goldan [:igoldan]]

:xidorn It seems like bug 1457912 caused some big displaylist_mutate regressions. They happen on Windows also.
Can you please take a look over this perf regressions and state whether we can fix it?

Comment 2

[Ionuț Goldan [:igoldan]]

Here are the gecko profiles:

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

after bug 1457912: https://perf-html.io/from-url/https%3A%2F%2Fqueue.taskcluster.net%2Fv1%2Ftask%2FQWkYrxGqSJqdaCB56qOq0Q%2Fruns%2F0%2Fartifacts%2Fpublic%2Ftest_info%2Fprofile_displaylist_mutate.zip

Comment 3

[Xidorn Quan [:xidorn] UTC+10]

So, I think the reason for this is likely that the scrollbars were not displayed due to that bug, and when that bug gets fixed, the scrollbars show up and slows down the test itself. I'm doing some try push to try proving that theory.

If that theory is true, we probably should just close this bug with WONTFIX because normal pages generally have scrollbars, so it makes more sense to show them in talos tests rather than hiding them purely for the sake of a seeming regression.

Comment 4

[Xidorn Quan [:xidorn] UTC+10]

These are four try pushes:
1) baseline (a m-c commit before bug 1457912):  https://treeherder.mozilla.org/#/jobs?repo=try&revision=ef0209e935de3fdb09b3d9f602fb4703fc249f94
2) with scrollbars explicitly enabled: https://treeherder.mozilla.org/#/jobs?repo=try&revision=d1752e7293d331598b002f5e1f14b1d9c0c462b8
3) with bug 1457912 applied: https://treeherder.mozilla.org/#/jobs?repo=try&revision=dce8e26497d95bf30dcb456ee0d57c2eb647fcbf
4) with bug 1457912 applied and scrollbars explicitly disabled: https://treeherder.mozilla.org/#/jobs?repo=try&revision=2623f630cf2737764682b82689ee714a3a69fc11

Comparison between (1) and (2): https://treeherder.mozilla.org/perf.html#/compare?originalProject=try&originalRevision=ef0209e935de3fdb09b3d9f602fb4703fc249f94&newProject=try&newRevision=d1752e7293d331598b002f5e1f14b1d9c0c462b8&framework=1
Comparison between (2) and (3): https://treeherder.mozilla.org/perf.html#/compare?originalProject=try&originalRevision=d1752e7293d331598b002f5e1f14b1d9c0c462b8&newProject=try&newRevision=dce8e26497d95bf30dcb456ee0d57c2eb647fcbf&framework=1
Comparison between (1) and (4): https://treeherder.mozilla.org/perf.html#/compare?originalProject=try&originalRevision=ef0209e935de3fdb09b3d9f602fb4703fc249f94&newProject=try&newRevision=2623f630cf2737764682b82689ee714a3a69fc11&framework=1

As can be seen from the comparison between (1) and (2), and (2) and (3), having scrollbars explicitly enabled has exactly the same impact as bug 1457912 on this test. And as can be seen from comparison between (1) and (4), with bug 1457912 and scrollbars explicitly disabled makes this impact disappear.

So this regression is basically just because of the bug fix itself which makes the scrollbars appear (as it should be), and thus I suggest we close this bug as WONTFIX since I think having scrollbars available should reflect real world cases better.

Comment 5

[Ionuț Goldan [:igoldan]]

(In reply to Xidorn Quan [:xidorn] UTC+10 from comment #4)
> These are four try pushes:
> 1) baseline (a m-c commit before bug 1457912): 
> https://treeherder.mozilla.org/#/
> jobs?repo=try&revision=ef0209e935de3fdb09b3d9f602fb4703fc249f94
> 2) with scrollbars explicitly enabled:
> https://treeherder.mozilla.org/#/
> jobs?repo=try&revision=d1752e7293d331598b002f5e1f14b1d9c0c462b8
> 3) with bug 1457912 applied:
> https://treeherder.mozilla.org/#/
> jobs?repo=try&revision=dce8e26497d95bf30dcb456ee0d57c2eb647fcbf
> 4) with bug 1457912 applied and scrollbars explicitly disabled:
> https://treeherder.mozilla.org/#/
> jobs?repo=try&revision=2623f630cf2737764682b82689ee714a3a69fc11
> 
> Comparison between (1) and (2):
> https://treeherder.mozilla.org/perf.html#/
> compare?originalProject=try&originalRevision=ef0209e935de3fdb09b3d9f602fb4703
> fc249f94&newProject=try&newRevision=d1752e7293d331598b002f5e1f14b1d9c0c462b8&
> framework=1
> Comparison between (2) and (3):
> https://treeherder.mozilla.org/perf.html#/
> compare?originalProject=try&originalRevision=d1752e7293d331598b002f5e1f14b1d9
> c0c462b8&newProject=try&newRevision=dce8e26497d95bf30dcb456ee0d57c2eb647fcbf&
> framework=1
> Comparison between (1) and (4):
> https://treeherder.mozilla.org/perf.html#/
> compare?originalProject=try&originalRevision=ef0209e935de3fdb09b3d9f602fb4703
> fc249f94&newProject=try&newRevision=2623f630cf2737764682b82689ee714a3a69fc11&
> framework=1
> 
> As can be seen from the comparison between (1) and (2), and (2) and (3),
> having scrollbars explicitly enabled has exactly the same impact as bug
> 1457912 on this test. And as can be seen from comparison between (1) and
> (4), with bug 1457912 and scrollbars explicitly disabled makes this impact
> disappear.
> 
> So this regression is basically just because of the bug fix itself which
> makes the scrollbars appear (as it should be), and thus I suggest we close
> this bug as WONTFIX since I think having scrollbars available should reflect
> real world cases better.

Thanks for the thorough check!